Project Settings Registry

Every configurable parameter of a RedAmon project, with what it means, what it accepts and whether an external agent may write it. GENERATED from recon_settings/registry.yaml; a test fails the build when this page and the registry disagree, so it cannot go stale.

There are 770 parameters. 702 are settable over the MCP surface at any time, 19 are fixed when a project is created, and 49 are not pipeline parameters at all. 14 of the settable ones are the engagement's LIMITS - a rate ceiling, an exclusion list, a time window, the agent's denylists - which the form and the MCP surface reach alike and which the pipeline enforces at scan start whatever tuning says.

For the narrative version with screenshots, see Project Settings Reference.

How to read this

Accepts is the ENFORCED bound, not a suggestion. A value outside it is refused by name rather than clamped, and one bad key refuses a whole batch.
Written vs resolved. Several values are corrected at scan start rather than refused: a rate above the engagement ceiling comes down to the ceiling, a container image outside the shipped set is pinned back to the default, and a wordlist path outside the project directory is dropped. Each is logged with a [guardrail] line.
Zero is not always the slowest value. Several rate fields treat 0 as UNLIMITED, which makes it the most aggressive setting available. Those say so.
Two levels. scanModules decides which PHASES run; a per-tool *Enabled flag decides which tools run inside a phase. Setting one without the other is a silent no-op. JS recon is the exception: it is no phase, jsReconEnabled alone switches it on, and it analyzes the JS files resource_enum and http_probe found, so it needs one of those. jsluice runs inside resource_enum.
One capability surface. Every settable field has an input in the project form and is writable through update_recon_settings; neither door reaches something the other cannot. The exception is the engagement RECORD, which is deliberately UI-only.

Agent (44)
AI attack-surface recon (15)
AlienVault OTX (2)
Amass (7)
Arjun (12)
BadDNS (5)
Banner grab (4)
Capture proxy (1)
Censys (2)
CriminalIP (2)
crt.sh (2)
CVE lookup (4)
CypherFix (6)
Denial of service (4)
DNS (4)
Domain recon AI hints (2)
Engagement (2)
ffuf (21)
Fireteam (8)
FOFA (3)
gau (21)
GitHub secret hunt (7)
GraphQL Cop (17)
GraphQL security (13)
GVM vulnerability scan (6)
HackerTarget (2)
hakrawler (11)
HTTP probe AI hints (5)
httpx (32)
Hydra (8)
Insecure deserialization (7)
IP attribution (2)
JS recon (32)
jsluice (14)
Katana (13)
Kiterunner (18)
knockpy (2)
masscan (9)
MITRE ATT&CK enrichment (7)
naabu (14)
Netlas (2)
Nmap (8)
Nuclei (30)
OpenAPI ingestion (6)
Origin discovery (9)
OSINT enrichment (1)
ParamSpider (4)
Path traversal (8)
Phishing (1)
Pipeline (4)
Pipeline AI master switch (2)
Port scan AI catalog (1)
Project identity (14)
puredns (6)
Remote code execution (4)
Resource enum AI hints (6)
Route exposure scan (22)
Rules of Engagement (40)
Security checks (37)
Serialized object scan (2)
Server-side request forgery (11)
Shodan (6)
SQL injection (3)
Subdomain discovery (1)
Subdomain takeover (8)
subfinder (3)
subjack (9)
Supply chain recon (4)
Supply chain scan (14)
Targeting and scope (11)
tlsx (18)
TruffleHog (16)
uncover (3)
urlscan.io (2)
Vhost / SNI enumeration (11)
VirusTotal (2)
WAF detection (2)
Wappalyzer (6)
Web cache poisoning (19)
WHOIS (2)
ZAP Ajax Spider (19)
ZoomEye (3)

Agent

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`agentActivatePostExplPhase`	true / false	`true`	settable	Whether the post-exploitation phase is available to the agent at all. A Rules of Engagement document may only set it to `false`.
`agentBindPortOnTarget`	1–65535 count	—	settable	Leave empty if unsure (agent will ask) A Rules of Engagement document may never set this.
`agentBruteForceMaxWordlistAttempts`	0–1000 count	`3`	settable	The attempt ceiling for one credential attack, which is what keeps it from running a whole wordlist against a live account.
`agentBruteforceSpeed`	0–10 count	`5`	settable	How aggressively the agent's credential attacks run, which trades attempts per second against lockout risk.
`agentChiselTunnelEnabled`	true / false	`false`	settable	Whether the agent may stand up a chisel tunnel, which gives it network reach beyond its own container. A Rules of Engagement document may only set it to `false`.
`agentCreateGraphImageOnInit`	true / false	`false`	settable	Renders a picture of the graph when a session starts, so the agent begins with a view of the estate.
`agentCypherMaxRetries`	0–10 count	`3`	settable	How many times agent retries a failed request before giving up on it.
`agentExecutionTraceMemorySteps`	1–10000 count	`100`	settable	Past steps kept in context
`agentExplSystemPrompt`	`free_text`	(empty)	settable	Injected during the exploitation phase. Leave empty for default. A Rules of Engagement document may never set this.
`agentGuardrailEnabled`	true / false	`true`	settable	The agent's own scope guardrail, which refuses a tool call aimed outside the project's targets. A Rules of Engagement document may only set it to `true`.
`agentInformationalSystemPrompt`	`free_text`	(empty)	settable	Injected during the informational phase. Leave empty for default. A Rules of Engagement document may never set this.
`agentKaliInstallAllowedPackages`	`free_text`	(empty)	settable	Comma-separated whitelist. If non-empty, ONLY these packages can be installed. A Rules of Engagement document may never set this.
`agentKaliInstallEnabled`	true / false	`false`	settable	Whether the agent may install additional packages inside its Kali sandbox. A Rules of Engagement document may only set it to `false`.
`agentKaliInstallForbiddenPackages`	`free_text`	(empty)	settable	Comma-separated blacklist. These packages must NEVER be installed.
`agentLatsBranching`	2–10 count	`6`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsEnabled`	true / false	`false`	settable	Whether agent runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`agentLatsMaxDepth`	2–10 depth	`6`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsMaxRollouts`	4–300 count	`50`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsMaxTreeNodes`	10–1000 count	`120`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsMinHypotheses`	2–4 count	`2`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsPhaseExploitation`	true / false	`true`	settable	Whether the LATS search runs during the exploitation phase as well as reconnaissance.
`agentLatsPhasePostExpl`	true / false	`false`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsPruneFloor`	0–1 ratio	`0.15`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLatsShadowMode`	true / false	`false`	settable	Runs the LATS search without acting on its choices, so its behaviour can be observed before it is trusted.
`agentLatsUctC`	0–10 ratio	`1.4`	settable	Part of the agent's LATS search configuration, which is how it explores several candidate actions before committing to one.
`agentLhost`	`free_text`	(empty)	settable	Your host machine's LAN IP that the target can reach (not the container's 172.x address). Leave empty for bind mode. A Rules of Engagement document may never set this.
`agentLlmParseMaxRetries`	0–20 count	`3`	settable	How many times agent retries a failed request before giving up on it.
`agentLogBackupCount`	0–100 count	`5`	settable	Rotated backups to keep
`agentLogMaxMb`	1–1024 count	`10`	settable	The size at which the agent's log file rolls, in megabytes.
`agentLport`	1–65535 port	—	settable	Leave empty for bind mode A Rules of Engagement document may never set this.
`agentMaxIterations`	1–1000 count	`100`	settable	LLM reasoning iterations limit
`agentNgrokTunnelEnabled`	true / false	`false`	settable	Whether the agent may expose a local listener through ngrok, which makes it reachable from the public internet. A Rules of Engagement document may only set it to `false`.
`agentOpenaiModel`	`free_text`	`claude-opus-4-6`	settable	Model used by the agent. Configure providers in Global Settings. A Rules of Engagement document may never set this.
`agentPayloadUseHttps`	true / false	`false`	settable	Whether a generated payload calls back over HTTPS rather than plain HTTP. A Rules of Engagement document may never set this.
`agentPlanMaxParallelTools`	1–50 count	`10`	settable	Concurrent tools per plan wave (root + fireteam); extras queue
`agentPostExplPhaseType`	`free_text`	`statefull`	settable	Stateful keeps Meterpreter/shell sessions between turns
`agentPostExplSystemPrompt`	`free_text`	(empty)	settable	Injected during the post-exploitation phase. Leave empty for default. A Rules of Engagement document may never set this.
`agentRequireApprovalForExploitation`	true / false	`true`	settable	Whether the agent must be approved before it moves from reconnaissance into exploitation. A Rules of Engagement document may only set it to `true`.
`agentRequireApprovalForPostExploitation`	true / false	`true`	settable	Whether the agent must be approved before it acts on a host it has already compromised. A Rules of Engagement document may only set it to `true`.
`agentRequireToolConfirmation`	true / false	`true`	settable	Whether every agent tool call waits for a human to approve it. A Rules of Engagement document may only set it to `true`.
`agentToolOutputMaxChars`	1000–100000 count	`40000`	settable	Truncation limit for tool output
`agentToolPhaseMap`	`json_object`	`{"query_graph":["informational","exploitation","post_exploitation"],"proxy_brain`	settable	Per-project overrides for which engagement phase each agent tool belongs to. A Rules of Engagement document may never set this.
`attackSkillConfig`	`attack_skill_config`	`{"builtIn":{"cve_exploit":true,"brute_force_credential_guess":false,"phishing_so`	settable	Per-project enable state of the attack skills: `builtIn` maps each built-in skill id to on or off, `user` maps the project owner's own uploaded skills by id. A user skill is on only when set to true (missing is off); at most 20 of the owner's skills may be on, ids the owner does not have are dropped, and a quarantined skill cannot be turned on. Rewritten across every project when a user deletes one of their own skills. A Rules of Engagement document may never set this.
`triageReviewBudget`	0–1000 count	`150`	settable	How many findings one triage run may have its evidence reviewed by the configured model before it stops and ranks the rest on the rules alone. 0 means no review and needs no model. A run never reviews more than 1000, whatever is stored.

AI attack-surface recon

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`aiSurfaceReconCacheEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconChatShapeProbeEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconJuliusProbePackEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconLatencyBaselineEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconMaxWorkers`	1–500 threads	`5`	settable	How many AI attack-surface recon operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `2`.
`aiSurfaceReconMcpHandshakeEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconMcpListToolsEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`aiSurfaceReconMcpYaraEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconModelListEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconOpenapiDiscoveryEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`aiSurfaceReconProbePackVersion`	`free_text`	`latest`	settable	Pinned probe-pack identifier recorded on every annotation (default "latest").
`aiSurfaceReconTimeout`	1–86400 seconds	`10`	settable	How long AI attack-surface recon may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`aiSurfaceReconUserAgent`	`free_text`	`RedAmon-AISurfaceRecon/1.0`	settable	User-Agent sent on every probe.
`aiSurfaceReconVectorDbReadEnabled`	true / false	`true`	settable	Whether AI attack-surface recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.

AlienVault OTX

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`otxEnabled`	true / false	`true`	settable	Whether otx runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`otxWorkers`	1–20 threads	`5`	settable	How many otx operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

Amass

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`amassActive`	true / false	`false`	settable	Amass active probing mode In stealth mode this is forced to `false`.
`amassBrute`	true / false	`false`	settable	Amass DNS brute-force In stealth mode this is forced to `false`.
`amassBruteWordlists`	`json_object`	`["default"]`	settable	Wordlists amass brute-forces with. Only the shipped list names are accepted.
`amassDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`caffix/amass:latest`	settable	The container image spawned for amass. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`amassEnabled`	true / false	`false`	settable	Whether amass runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`amassMaxResults`	1–50000 count	`50000`	settable	Maximum subdomains taken from an amass run. In stealth mode this is lowered to at most 100; a quieter value is kept.
`amassTimeout`	1–120 minutes	`10`	settable	How long amass may run, in MINUTES rather than seconds. The name reads like every other timeout in the model and the unit does not match it.

Arjun

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`arjunChunkSize`	10–5000 count	`500`	settable	Parameters tested per request batch. Lower = more requests, higher accuracy
`arjunCustomHeaders`	`http_header`	(empty)	settable	Add auth tokens or custom headers for authenticated parameter testing
`arjunDisableRedirects`	true / false	`false`	settable	Disable following redirects
`arjunEnabled`	true / false	`true`	settable	Whether arjun runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`arjunMaxEndpoints`	1–50000 count	`50000`	settable	The most results arjun accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`arjunMethods`	`free_text`	`["GET","POST"]`	settable	HTTP methods: ["GET", "POST", "PUT", "DELETE", "PATCH"]
`arjunPassive`	true / false	`false`	settable	Passive mode (no requests) In stealth mode this is forced to `true`.
`arjunRateLimit`	0–100000 rps	`0`	settable	Requests per second arjun sends while probing for parameters. ZERO MEANS UNLIMITED, so 0 is the most aggressive value available and not the safest. Under an engagement ceiling a 0 here is rewritten to the ceiling at scan start. 0 means UNLIMITED here, so it is the fastest value and not the safest. Capped to the engagement rate ceiling at scan start.
`arjunScanTimeout`	60–86400 seconds	`600`	settable	How long arjun may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`arjunStable`	true / false	`false`	settable	Stability mode (slower but more reliable)
`arjunThreads`	1–20 threads	`2`	settable	How many arjun operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`arjunTimeout`	1–86400 seconds	`15`	settable	How long arjun may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

BadDNS

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`baddnsDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`redamon-baddns:latest`	settable	The container image spawned for baddns. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`baddnsEnabled`	true / false	`false`	settable	Whether baddns runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`baddnsModules`	`free_text`	`["cname","ns","mx","txt","spf"]`	settable	Subset of BadDNS modules to run. Valid: cname, ns, mx, txt, spf, dmarc, wildcard, nsec, references, zonetransfer. (MTA-STS exists in baddns 2.1.0 but is not CLI-addressable due to an upstream validator regex bug -- omit.) Default: ["cname","ns","mx","txt","spf"]
`baddnsNameservers`	`free_text`	(empty)	settable	Optional custom DNS resolvers (e.g. ["1.1.1.1","8.8.8.8"]). Empty = system resolvers. A Rules of Engagement document may never set this.
`baddnsRunTimeout`	1–86400 seconds	`1800`	settable	How long baddns may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

phase port_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`bannerGrabEnabled`	true / false	`true`	settable	Whether Banner grab runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`bannerGrabMaxLength`	100–5000 bytes	`1000`	settable	Truncate banners longer than this (chars)
`bannerGrabThreads`	1–500 threads	`20`	settable	How many Banner grab operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`bannerGrabTimeout`	1–86400 seconds	`5`	settable	How long Banner grab may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

Capture proxy

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`captureProxyEnabled`	true / false	`false`	settable	Whether the HTTP capture proxy records the scan's traffic. Capture runs off the scan's critical path, so a capture failure never fails the scan.

Censys

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`censysEnabled`	true / false	`false`	settable	Whether censys runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`censysWorkers`	1–20 threads	`5`	settable	How many censys operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

CriminalIP

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`criminalIpEnabled`	true / false	`false`	settable	Whether CriminalIP runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`criminalIpWorkers`	1–20 threads	`5`	settable	How many CriminalIP operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

crt.sh

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`crtshEnabled`	true / false	`true`	settable	Whether crtsh runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`crtshMaxResults`	1–50000 count	`5000`	settable	Maximum certificate-transparency records read from crt.sh in one pass. In stealth mode this is lowered to at most 100; a quieter value is kept.

CVE lookup

phase vuln_scan · passive traffic

Field	Accepts	Default	MCP	Meaning
`cveLookupEnabled`	true / false	`true`	settable	Whether CVE lookup runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`cveLookupMaxCves`	1–100 count	`20`	settable	Maximum CVEs attached to one service before enrichment stops adding more.
`cveLookupMinCvss`	0–10 ratio	`0`	settable	Minimum CVSS score (0-10)
`cveLookupSource`	`free_text`	`nvd`	settable	Which vulnerability database CVE enrichment queries.

CypherFix

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`cypherfixBranchPrefix`	`free_text`	`cypherfix/`	settable	Prefix for fix branch names (e.g., cypherfix/rem-abc123). A Rules of Engagement document may never set this.
`cypherfixDefaultBranch`	`free_text`	`main`	settable	Base branch for creating fix branches (default: main). A Rules of Engagement document may never set this.
`cypherfixDefaultRepo`	`free_text`	(empty)	settable	GitHub repository to fix (owner/repo format). Can be overridden per remediation. A Rules of Engagement document may never set this.
`cypherfixGithubToken`	`free_text`	(empty)	closed	Personal access token with Withheld from every read on the MCP surface.
`cypherfixLlmModel`	`free_text`	(empty)	settable	No longer used. The Triage review and CodeFix models are chosen per user, for all of that user's projects, in Global Settings > LLM Providers > Models by feature. Kept only so older exports and presets still load. A Rules of Engagement document may never set this.
`cypherfixRequireApproval`	true / false	`true`	settable	Whether a CypherFix patch must be approved by a human before it is pushed.

Denial of service

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`dosAssessmentOnly`	true / false	`false`	settable	Only check for availability vulnerabilities (nmap scripts, nuclei) without active testing.
`dosConcurrentConnections`	10–10000 threads	`1000`	settable	How many dos operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`dosMaxAttempts`	1–10 count	`3`	settable	Max different vectors to try before reporting service is resilient.
`dosMaxDuration`	10–300 seconds	`60`	settable	The longest one denial-of-service attempt may run, in seconds. It is only reachable at all when roeAllowDos permits the technique.

DNS

phase domain_discovery · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`dnsEnabled`	true / false	`true`	settable	Whether dns runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`dnsMaxRetries`	1–10 count	`3`	settable	How many times dns retries a failed request before giving up on it.
`dnsMaxWorkers`	1–200 threads	`80`	settable	How many dns operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `5`.
`dnsRecordParallelism`	true / false	`true`	settable	How many dns operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `false`.

Domain recon AI hints

phase domain_discovery · sends no traffic

Field	Accepts	Default	MCP	Meaning
`domainReconAiNsHintEnabled`	true / false	`true`	settable	Whether Domain recon AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`domainReconAiTxtHintEnabled`	true / false	`true`	settable	Whether Domain recon AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.

Engagement

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`engagementIdentityHeader`	`http_header`	(empty)	settable	A header every request carries so the target's operators can attribute the traffic to this engagement, for example 'X-Bug-Bounty: your-handle'. Empty sends none. Many programs require one, and it is the difference between a scan that gets a question and one that gets a block. Validated like any other header: no CR or LF, and not Host, Authorization, Cookie or Proxy-*.
`engagementKind`	`internal`, `third_party`	`internal`	create-only	Who the target belongs to, and therefore what has to be true before a scan may start. 'internal' is your own estate. 'third_party' is somebody else's, and such a project MUST carry a non-zero rate ceiling and an authorization record or start_recon refuses it. Set at creation and immutable afterwards: converting a project after the fact would either claim an authority nobody granted or drop a ceiling a human put there. Every project created before this field existed reads as 'internal', so an old project with no ceiling is flagged rather than blocked. Fixed at project creation; refused on an existing project.

ffuf

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`ffufAiExtensions`	true / false	`false`	settable	When true, FFuf file extensions are picked per-target by AI from response headers (requires aiInPipeline=true). The static ffufExtensions list is ignored.
`ffufAiUseJev`	true / false	`false`	settable	Engine for the FFuf extension picker: false uses the LLM in aiPipelineModel, true uses TypeSafe AI Jev (model jev-1.13.0). Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one the hook uses its static fallback. Jev picks only from a fixed 40-extension catalog.
`ffufAutoCalibrate`	true / false	`true`	settable	Learns what a 'not found' response looks like on this target before filtering, which removes most soft-404 noise.
`ffufCustomHeaders`	`http_header`	(empty)	settable	One header per line. Sent with every request
`ffufEnabled`	true / false	`false`	settable	Whether ffuf runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`ffufExtensions`	`free_text`	(empty)	settable	File extensions to fuzz, e.g. [".php", ".asp"]
`ffufFilterCodes`	`status_codes`	(empty)	settable	HTTP status codes to filter out
`ffufFilterSize`	`free_text`	(empty)	settable	Filter responses by size
`ffufFollowRedirects`	true / false	`false`	settable	Follows a redirect rather than recording it, which can turn one request into several.
`ffufJevBasePaths`	true / false	`false`	settable	Picks the smart-fuzz base paths with TypeSafe AI Jev (model jev-1.13.0): false keeps the random pick that fills ffufSmartFuzzMaxBasePaths, true asks Jev, per discovered directory, whether it is likely to hold sensitive, administrative or application content, and FFuf fuzzes the directories Jev ranks highest. Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one the random pick is used. Jev only ranks the directories the crawl found, and the cap keeps the same number either way. Jev's pick is kept in the recon output next to the deterministic one.
`ffufMatchCodes`	`status_codes`	`[200,201,204,301,302,307,308,401,403,405]`	settable	HTTP status codes to match, e.g. [200, 301, 403]
`ffufMaxTime`	60–86400 seconds	`1800`	settable	Max total time in seconds
`ffufParallelism`	1–50 threads	`20`	settable	How many ffuf operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`ffufRate`	0–100000 rps	`0`	settable	Requests per second ffuf sends. ZERO MEANS UNLIMITED, so 0 is the most aggressive value available and not the safest, and it is also the shipped default. Under an engagement ceiling a 0 here is rewritten to the ceiling at scan start. 0 means UNLIMITED here, so it is the fastest value and not the safest. Capped to the engagement rate ceiling at scan start.
`ffufRecursion`	true / false	`false`	settable	Follows a discovered directory down into its own contents, which multiplies the request count.
`ffufRecursionDepth`	1–5 depth	`2`	settable	How many directories deep recursion goes before it stops.
`ffufSmartFuzz`	true / false	`true`	settable	Picks extensions and wordlist entries from what the target has already revealed rather than fuzzing blind.
`ffufSmartFuzzMaxBasePaths`	1–50 count	`20`	settable	The most crawler-discovered directory prefixes smart fuzz runs the whole wordlist under, on every target. Each one adds a full wordlist pass per target, so this multiplies the request count. When crawlers find more, a random sample of this size is fuzzed.
`ffufThreads`	1–200 threads	`40`	settable	How many ffuf operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`ffufTimeout`	1–86400 seconds	`10`	settable	How long ffuf may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`ffufWordlist`	`project_file`	`/usr/share/seclists/Discovery/Web-Content/common.txt`	settable	Path to wordlist file

Fireteam

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`fireteamAllowedPhases`	`free_text`	`["informational","exploitation","post_exploitation"]`	settable	Engagement phases in which the fireteam may operate. A Rules of Engagement document may never set this.
`fireteamConfirmationTimeoutSec`	0–86400 count	`600`	settable	How long a fireteam member waits for confirmation before giving up on a step.
`fireteamEnabled`	true / false	`true`	settable	Whether the agent may split work across a fireteam of sub-agents.
`fireteamMaxConcurrent`	1–8 count	`5`	settable	1-8. Upper limit on members in-flight at once.
`fireteamMaxMembers`	2–8 count	`5`	settable	2-8. Hard cap on fireteam size the LLM can request.
`fireteamMemberMaxIterations`	5–50 count	`10`	settable	5-50. Each member's ReAct budget before it exits.
`fireteamPropensity`	1–5 count	`3`	settable	How strongly the agent leans toward deploying a fireteam over single-agent or plan_tools. Injected into the system prompt as a directive the LLM must follow.
`fireteamTimeoutSec`	60–7200 count	`7200`	settable	60-7200. Hard wall-clock ceiling for the whole fireteam.

FOFA

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`fofaEnabled`	true / false	`false`	settable	Whether fofa runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`fofaMaxResults`	1–10000 count	`1000`	settable	The most results fofa accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here. An empty value is replaced by the default rather than honoured.
`fofaWorkers`	1–20 threads	`5`	settable	How many fofa operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

gau

phase resource_enum · passive traffic

Field	Accepts	Default	MCP	Meaning
`gauBlacklistExtensions`	`free_text`	`["png","jpg","jpeg","gif","svg","ico","webp","avif","css","woff","woff2","ttf","`	settable	File extensions to skip, e.g. [".jpg", ".css"]
`gauDetectMethods`	true / false	`true`	settable	Detect allowed HTTP methods
`gauDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`sxcurity/gau:latest`	settable	The container image spawned for gau. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`gauEnabled`	true / false	`false`	settable	Whether gau runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `true`.
`gauFilterDeadEndpoints`	true / false	`true`	settable	Drops archived URLs that no longer respond, so the result reflects the live surface.
`gauMaxUrls`	1–100000 count	`50000`	settable	The most results gau accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`gauMethodDetectRateLimit`	1–100000 rps	`50`	settable	Requests per second the gau method-detection pass sends to the target. Active traffic despite gau's passive reputation, so the engagement ceiling caps it. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `2`.
`gauMethodDetectThreads`	1–500 threads	`25`	settable	How many gau operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`gauMethodDetectTimeout`	1–86400 seconds	`5`	settable	How long gau may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`gauProviders`	`free_text`	`["wayback","commoncrawl","otx","urlscan"]`	settable	Sources: "wayback", "commoncrawl", "otx", "urlscan"
`gauThreads`	1–20 threads	`5`	settable	How many gau operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`gauTimeout`	10–86400 seconds	`60`	settable	How long gau may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`gauVerbose`	true / false	`false`	settable	Emits per-URL progress, which is useful while debugging and noisy otherwise.
`gauVerifyAcceptStatus`	`status_codes`	`[200,201,301,302,307,308,401,403]`	settable	Status codes that indicate a live URL. Include 401/403 for auth-protected endpoints
`gauVerifyDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/httpx:latest`	settable	The container image spawned for gau. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`gauVerifyRateLimit`	1–100000 rps	`100`	settable	Requests per second the gau verification pass sends. gau itself only reads public archives, but verification dials the target, so this is active traffic and the engagement ceiling caps it. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `2`.
`gauVerifyThreads`	1–100 threads	`50`	settable	How many gau operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`gauVerifyTimeout`	1–86400 seconds	`5`	settable	How long gau may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`gauVerifyUrls`	true / false	`true`	settable	Verify discovered URLs are alive
`gauWorkers`	1–20 threads	`10`	settable	How many gau operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`gauYearRange`	`free_text`	(empty)	settable	Filter Wayback Machine by year (e.g., 2020, 2024)

GitHub secret hunt

standalone · passive traffic

Field	Accepts	Default	MCP	Meaning
`githubMaxCommits`	1–1000 count	`100`	settable	The most results github accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`githubOutputJson`	true / false	`true`	settable	Emits machine-readable output alongside the human-readable report.
`githubScanCommits`	true / false	`true`	settable	Walks commit history as well as the working tree, which is where most leaked secrets actually live.
`githubScanGists`	true / false	`true`	settable	Includes the organisation members' public gists in the hunt.
`githubScanMembers`	true / false	`false`	settable	Includes the organisation members' own repositories in the hunt.
`githubTargetOrg`	`github_name`	(empty)	create-only	The GitHub organisation the secret hunt scans. It points a scanner at someone else's assets, so it is scope by another name: set when the project is created, and afterwards changed only by the project form or by update_project_scope under its own permission. Fixed at project creation; refused on an existing project.
`githubTargetRepos`	`github_repo_list`	(empty)	create-only	Comma-separated list. Leave empty to scan all repositories. Fixed at project creation; refused on an existing project.

GraphQL Cop

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`graphqlCopDebug`	true / false	`false`	settable	-d flag: adds X-GraphQL-Cop-Test header per request
`graphqlCopDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`dolevf/graphql-cop:1.14`	settable	The container image spawned for GraphQL Cop. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`graphqlCopEnabled`	true / false	`false`	settable	Whether GraphQL Cop runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`graphqlCopForceScan`	true / false	`false`	settable	-f flag: scan even if endpoint doesn't look GraphQL-like
`graphqlCopTestAliasOverloading`	true / false	`true`	settable	HIGH DoS, 101-alias rate-limit bypass (default true, disabled in stealth) In stealth mode this is forced to `false`.
`graphqlCopTestBatchQuery`	true / false	`true`	settable	HIGH DoS, array-based batch queries (default true, disabled in stealth) In stealth mode this is forced to `false`.
`graphqlCopTestCircularIntrospection`	true / false	`true`	settable	HIGH DoS, deep nested introspection (default true, disabled in stealth) In stealth mode this is forced to `false`.
`graphqlCopTestDirectiveOverloading`	true / false	`true`	settable	HIGH DoS (default true, disabled in stealth) In stealth mode this is forced to `false`.
`graphqlCopTestFieldSuggestions`	true / false	`true`	settable	LOW info-leak check (default true)
`graphqlCopTestGetMethod`	true / false	`true`	settable	MEDIUM CSRF, GET-method queries allowed (default true)
`graphqlCopTestGetMutation`	true / false	`true`	settable	MEDIUM CSRF, GET-based mutations (default true)
`graphqlCopTestGraphiql`	true / false	`true`	settable	LOW info-leak, detects GraphiQL/Playground UI (default true)
`graphqlCopTestIntrospection`	true / false	`false`	settable	HIGH info-leak check (default false, dedupes with native scanner)
`graphqlCopTestPostCsrf`	true / false	`true`	settable	MEDIUM CSRF, url-encoded POST (default true)
`graphqlCopTestTraceMode`	true / false	`true`	settable	INFO, Apollo tracing extension disclosure (default true)
`graphqlCopTestUnhandledError`	true / false	`true`	settable	INFO info-leak, exception stack traces (default true)
`graphqlCopTimeout`	10–600 seconds	`120`	settable	How long GraphQL Cop may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

GraphQL security

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`graphqlAuthHeader`	`free_text`	(empty)	settable	Header name when graphqlAuthType = "custom" A Rules of Engagement document may never set this.
`graphqlAuthType`	`free_text`	(empty)	settable	"", "bearer", "basic", "cookie", "custom"
`graphqlAuthValue`	`free_text`	(empty)	settable	Auth credential (token / user:pass / cookie string) A Rules of Engagement document may never set this. Withheld from every read on the MCP surface.
`graphqlConcurrency`	1–20 threads	`5`	settable	How many graphql operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`graphqlDepthLimit`	1–50 depth	`10`	settable	Max introspection query nesting depth (default 10)
`graphqlEndpoints`	`free_text`	(empty)	settable	Comma-separated custom GraphQL endpoints (empty = auto-discover) A Rules of Engagement document may never set this.
`graphqlIntrospectionTest`	true / false	`true`	settable	Probe __schema to detect exposed introspection (default true) In stealth mode this is forced to `true`.
`graphqlRateLimit`	0–100 rps	`10`	settable	Requests per second graphql sends. Capped to the engagement ceiling at scan start, so a value above the ceiling is rewritten rather than refused. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `2`.
`graphqlRetryBackoff`	0–60 ratio	`2`	settable	Exponential backoff base seconds between retries (default 2.0)
`graphqlRetryCount`	0–10 count	`3`	settable	How many times graphql retries a failed request before giving up on it.
`graphqlSecurityEnabled`	true / false	`false`	settable	Whether graphql runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `true`.
`graphqlTimeout`	1–600 seconds	`30`	settable	How long graphql may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan. In stealth mode this is forced to `60`.
`graphqlVerifySsl`	true / false	`true`	settable	Reject invalid / self-signed TLS certs (default true)

GVM vulnerability scan

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`gvmCleanupAfterScan`	true / false	`true`	settable	Removes the GVM task and its target after the scan, which keeps the appliance from accumulating state.
`gvmPollInterval`	5–300 count	`30`	settable	Seconds between scan status checks. Lower values give faster log updates.
`gvmPortList`	`free_text`	`All TCP and Nmap top 100 UDP`	settable	Which ports GVM probes on every target. The full IANA UDP range is the single biggest driver of scan time: closed UDP ports answer with silence and must each be timed out, and GVM reports no progress percentage until that finishes.
`gvmScanConfig`	`free_text`	`Full and fast`	settable	GVM scan configuration preset. “Full and fast” is recommended for most targets.
`gvmScanTargets`	`both`, `ips_only`, `hostnames_only`	`both`	create-only	Which targets from recon data to scan: both, ips_only or hostnames_only. Both provides the most thorough coverage. An unknown value makes the GVM scan select nothing, so the value set is closed. Fixed at project creation; refused on an existing project.
`gvmTaskTimeout`	0–86400 seconds	`14400`	settable	How long gvm may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

HackerTarget

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`hackerTargetEnabled`	true / false	`true`	settable	Whether hackertarget runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`hackerTargetMaxResults`	1–50000 count	`5000`	settable	Maximum subdomains taken from the HackerTarget API in one pass. In stealth mode this is lowered to at most 100; a quieter value is kept.

hakrawler

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`hakrawlerCustomHeaders`	`http_header`	(empty)	settable	One header per line (e.g., Cookie: value). Sent with every request
`hakrawlerDepth`	1–10 depth	`2`	settable	How many links deep to follow
`hakrawlerDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`jauderho/hakrawler:latest`	settable	The container image spawned for hakrawler. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`hakrawlerEnabled`	true / false	`true`	settable	Whether hakrawler runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`hakrawlerIncludeSubs`	true / false	`true`	settable	Follows links onto subdomains of the target rather than staying on one host.
`hakrawlerInsecure`	true / false	`true`	settable	Skip TLS verification
`hakrawlerJevSeedOrder`	true / false	`false`	settable	Orders Hakrawler's seed URLs with TypeSafe AI Jev (model jev-1.13.0): false keeps the alphabetical order, true asks Jev, per probed host, whether it is likely to have a rich web application surface, and Hakrawler crawls the promising hosts first, so under hakrawlerMaxUrls they are reached before the cap. Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one the alphabetical order is used. Ordering only: every seed stays in the list, and partial recon keeps the alphabetical order. Jev's order is kept in the recon output next to the alphabetical one.
`hakrawlerMaxUrls`	1–100000 count	`50000`	settable	The most results hakrawler accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`hakrawlerParallelism`	1–10 threads	`5`	settable	How many hakrawler operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`hakrawlerThreads`	1–20 threads	`5`	settable	How many hakrawler operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. Capped to the engagement rate ceiling at scan start.
`hakrawlerTimeout`	5–86400 seconds	`30`	settable	How long hakrawler may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

HTTP probe AI hints

phase http_probe · sends no traffic

Field	Accepts	Default	MCP	Meaning
`httpProbeAiFaviconHashEnabled`	true / false	`true`	settable	Whether HTTP probe AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`httpProbeAiHeaderScanEnabled`	true / false	`true`	settable	Whether HTTP probe AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`httpProbeAiTitleDetectionEnabled`	true / false	`true`	settable	Whether HTTP probe AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`httpProbeAiWappalyzerEnabled`	true / false	`true`	settable	Whether HTTP probe AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`httpxJevPageType`	true / false	`false`	settable	Labels each probed URL with TypeSafe AI Jev (model jev-1.13.0): false labels nothing, true asks Jev whether a page is a login wall, a parked domain, a default install page, a placeholder or an error page, and calls it an app otherwise. A deterministic pre-filter supplies the label for any page Jev gives no answer for. Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one only the pre-filter labels. The label is written onto the page's Endpoint in the graph (page_class) and changes nothing about what is scanned. No traffic to the target: at most 300 distinct pages per scan, within 60 seconds.

httpx

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`httpxCustomHeaders`	`http_header`	`["User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTM`	settable	Extra headers sent with every httpx request. Validated: a header may not contain CR or LF, and Host, Authorization, Cookie and Proxy-* are refused, because each would change where the request goes or what it carries rather than annotating it.
`httpxDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/httpx:latest`	settable	The container image spawned for httpx. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`httpxEnabled`	true / false	`true`	settable	Whether httpx runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`httpxFilterCodes`	`status_codes`	(empty)	settable	Hide responses with these status codes
`httpxFollowRedirects`	true / false	`true`	settable	Follows a redirect to its destination rather than recording the redirect itself.
`httpxIncludeResponse`	true / false	`true`	settable	Store full response body
`httpxIncludeResponseHeaders`	true / false	`true`	settable	Records the full response headers, which later checks read and which make the output much larger.
`httpxMatchCodes`	`status_codes`	(empty)	settable	Only show responses with these status codes
`httpxMaxRedirects`	1–50 count	`10`	settable	Maximum redirect chain depth to prevent loops
`httpxPaths`	`free_text`	(empty)	settable	Additional URL paths to probe
`httpxProbeAsn`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeCdn`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeCname`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeContentLength`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeContentType`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeFavicon`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is. In stealth mode this is forced to `false`.
`httpxProbeHash`	`free_text`	`sha256`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeIp`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeJarm`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is. In stealth mode this is forced to `false`.
`httpxProbeLineCount`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeResponseTime`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeServer`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeStatusCode`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeTechDetect`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeTitle`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeTlsGrab`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeTlsInfo`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxProbeWordCount`	true / false	`true`	settable	Whether httpx records this attribute while probing. Each probe adds work per host, and several add requests beyond the first, so the set of probes decides both how much the scan learns and how loud it is.
`httpxRateLimit`	1–100000 rps	`50`	settable	Requests per second httpx sends while probing. Capped by the engagement ceiling at scan start. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `2`.
`httpxRetries`	0–10 count	`2`	settable	How many times httpx retries a failed request before giving up on it.
`httpxThreads`	1–200 threads	`50`	settable	How many httpx operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`httpxTimeout`	1–86400 seconds	`10`	settable	How long httpx may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

Hydra

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`hydraConnectionTimeout`	5–120 seconds	`32`	settable	How long hydra may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`hydraEnabled`	true / false	`true`	settable	Whether hydra runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`hydraExtraChecks`	`free_text`	`nsr`	settable	Additional protocol-specific checks run alongside the credential attempts.
`hydraMaxWordlistAttempts`	1–10 count	`3`	settable	How many wordlist strategies to try before giving up.
`hydraStopOnFirstFound`	true / false	`true`	settable	Stop immediately when valid credentials are found.
`hydraThreads`	1–64 threads	`16`	settable	How many hydra operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`hydraVerbose`	true / false	`true`	settable	Show each login attempt in output for progress tracking.
`hydraWaitBetweenConnections`	0–300 threads	`0`	settable	How many hydra operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.

Insecure deserialization

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`deserializationExecGadgetsEnabled`	true / false	`false`	settable	Whether the Insecure Deserialization skill may escalate a confirmed sink to one code-execution gadget (Step 7). Off by default. Its proof travels over the out-of-band callback, so it also needs the OOB callback on; without it the Step 7 block is left out of the prompt and the agent stops at the confirmed sink.
`deserializationFindSinksEnabled`	true / false	`true`	settable	Whether the Insecure Deserialization skill may sweep the target's inputs for sinks recon never flagged. When off, the agent only confirms the candidates the serialized-object scan wrote to the graph.
`deserializationOobCallbackEnabled`	true / false	`true`	settable	Whether the Insecure Deserialization skill may confirm a sink with the out-of-band interactsh oracle, which makes the target look up a domain registered on the OOB provider. When off, the setup block is left out of the prompt and the agent confirms through the error and timing channels only, never through external callback infrastructure.
`deserializationOobProvider`	`hostname`	`oast.fun`	settable	interactsh server the deserialization oracle registers its callback domain on. Use a self-hosted instance if oast.fun is blocked or a third-party callback service is not allowed. Empty means oast.fun. Only used when the OOB callback is on. A Rules of Engagement document may never set this.
`deserializationPharEnabled`	true / false	`false`	settable	Whether the Insecure Deserialization skill may use PHP PHAR polyglots, which means uploading a crafted file to the target. Off by default. Only used when php is among the runtimes.
`deserializationRuntimes`	`free_text`	`java,java_typed,python,php,node,ruby,dotnet`	settable	Comma-separated runtimes whose oracle blocks ship in the Insecure Deserialization prompt: java, java_typed, python, php, node, ruby, dotnet. Unknown names are ignored; an empty list ships every runtime.
`deserializationTimingEnabled`	true / false	`true`	settable	Whether the Insecure Deserialization skill may confirm a sink through response timing, which holds a request open on the target until a delay or timeout ends. Turn it off on fragile targets or when the rules of engagement forbid anything that ties up a server worker.

IP attribution

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`ipAttributionEnabled`	true / false	`true`	settable	Whether IP attribution runs. After the HTTP probe it tags every IP of the run shared_edge (a CDN edge serving many tenants), dedicated (the certificate it presents without SNI covers a target domain) or unknown, never "not ours". Vhost/SNI skips its wordlist on a shared edge, origin discovery never reports one as an origin, Shodan and Netlas write no CVE on one, and the Priority Board caps the reach of findings on one. Off, IPs keep no tag and those consumers fall back to their own CDN checks.
`ipAttributionTlsMaxProbes`	0–5000 count	`500`	settable	The most TLS handshakes without SNI IP attribution makes in one run, one per IP where port 443 was found open, to read the certificate the IP falls back to. 0 means no handshake at all: the tag then uses only the CDN signals the run already holds. The handshakes skip RoE-excluded hosts and are paced at the engagement rate ceiling, and stealth mode sets this to 0. In stealth mode this is forced to `0`.

JS recon

phase js_recon · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`jsReconAiSdkDetectionEnabled`	true / false	`true`	settable	Whether JS recon looks for AI SDKs in the JS it analyzes: LLM and vector database SDK imports, hard-coded provider keys, browser-mode opt-ins and AI front-end markers. Sends no extra traffic. Does nothing while jsReconEnabled is off.
`jsReconConcurrency`	1–30 threads	`10`	settable	How many JS recon operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`jsReconCustomEndpointKeywords`	`free_text`	(empty)	closed	The uploaded custom endpoint-keyword list. Written only by the custom-files endpoint. Written only by /api/js-recon/[projectId]/custom-files.
`jsReconCustomFrameworks`	`free_text`	(empty)	closed	The uploaded custom framework-signature list. Written only by the custom-files endpoint. Written only by /api/js-recon/[projectId]/custom-files.
`jsReconCustomPackages`	`free_text`	(empty)	closed	The uploaded custom package list. Written only by the custom-files endpoint. Written only by /api/js-recon/[projectId]/custom-files.
`jsReconCustomPatterns`	`free_text`	(empty)	closed	The uploaded custom secret-pattern file. Written only by the custom-files endpoint, which also places it on disk. Written only by /api/js-recon/[projectId]/custom-files.
`jsReconCustomSourcemapPaths`	`free_text`	(empty)	closed	The uploaded custom source-map path list. Written only by the custom-files endpoint. Written only by /api/js-recon/[projectId]/custom-files.
`jsReconDependencyCheck`	true / false	`true`	settable	Checks the packages a bundle names against known-vulnerable versions.
`jsReconDevComments`	true / false	`true`	settable	Extract developer comments
`jsReconDomSinks`	true / false	`true`	settable	Detect DOM XSS sinks
`jsReconEnabled`	true / false	`false`	settable	Whether JS recon runs. This flag alone switches it on: JS recon is not a scanModules phase. It runs after resource enumeration and analyzes the JS files the crawlers and HTTP probing found (and any uploaded files), so with neither resource_enum nor http_probe in scanModules it has nothing to read.
`jsReconEndpointAcceptStatus`	`status_codes`	`[200,201,204,301,302,307,308,401,403,405]`	settable	HTTP status codes treated as live during endpoint verification An empty value is replaced by the default rather than honoured.
`jsReconEndpointConcurrency`	1–20 threads	`10`	settable	How many JS recon operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`jsReconEndpointCustomHeaders`	`http_header`	(empty)	settable	Header lines used when verifying extracted endpoints
`jsReconExtractEndpoints`	true / false	`true`	settable	Extracts API endpoints referenced in JavaScript, which is the main reason to read it at all.
`jsReconFrameworkDetect`	true / false	`true`	settable	Identifies the framework a bundle was built with, which decides which later checks apply.
`jsReconIncludeArchivedJs`	true / false	`true`	settable	Includes JavaScript recovered from web archives, which can reveal endpoints the live site no longer exposes.
`jsReconIncludeChunks`	true / false	`true`	settable	Includes lazily-loaded chunks, which multiplies the file count on a modern bundler's output. In stealth mode this is forced to `false`.
`jsReconIncludeFrameworkJs`	true / false	`true`	settable	Includes framework bundles, which are large, mostly third-party, and occasionally where a key is embedded. In stealth mode this is forced to `false`.
`jsReconMaxFiles`	10–10000 count	`10000`	settable	The most JS files JS recon downloads and analyzes in one run, shared round-robin between hosts; files over the cap are recorded as a coverage gap. The memory governor budgets it against available RAM, so the effective value can be below what is written here. In stealth mode this is forced to `50`.
`jsReconMinConfidence`	`free_text`	`low`	settable	"low", "medium", "high"
`jsReconRegexPatterns`	true / false	`true`	settable	Additional regular expressions applied to retrieved JavaScript, on top of the shipped set.
`jsReconSourceMaps`	true / false	`true`	settable	Retrieves source maps when they are published, which recovers the original source rather than the bundle.
`jsReconStandaloneCrawlDepth`	0–50 depth	`3`	settable	How many links deep a standalone JS recon run crawls from its seed.
`jsReconStandaloneCrawlScope`	`free_text`	`subdomain`	settable	How far a standalone JS recon run may crawl from its seed: same host, same domain, or anywhere.
`jsReconTimeout`	60–86400 seconds	`900`	settable	How long JS recon may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`jsReconTrufflehogEnabled`	true / false	`true`	settable	Runs TruffleHog over local copies of the JavaScript JS recon analyses, in an isolated container, and merges its findings with JS recon's: one finding per key. With key validation on, TruffleHog also checks the keys it finds with their providers (never a host named in the JavaScript); with it off the container has no network.
`jsReconTrufflehogResults`	`verified`, `unknown`, `unverified`	`["verified","unknown","unverified"]`	settable	Which keys found only by TruffleHog are kept when key validation is on: verified (the provider accepted it), unknown (the check errored) and unverified (not shown live, or not checkable). A key JS recon also found is always kept. Ignored when key validation is off; empty means all three. An empty value is replaced by the default rather than honoured.
`jsReconUploadedFiles`	`free_text`	(empty)	closed	JavaScript files uploaded for analysis rather than fetched from the target. Written only by the upload endpoint, which also places each file on disk. Written only by /api/js-recon/[projectId]/upload.
`jsReconValidateEndpoints`	true / false	`false`	settable	Probe extracted endpoints and drop confirmed-dead ones before graph ingestion (default off)
`jsReconValidateKeys`	true / false	`true`	settable	Validate discovered API keys by sending each one to its provider (and, with TruffleHog on, letting TruffleHog check the keys it finds the same way): a live key is confirmed, but every key found is used against a third-party service. A Rules of Engagement document can only turn it off. A Rules of Engagement document may only set it to `false`. In stealth mode this is forced to `false`.
`jsReconValidationTimeout`	1–30 seconds	`5`	settable	How long JS recon may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

jsluice

phase js_recon · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`jsluiceConcurrency`	1–20 threads	`5`	settable	How many jsluice operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`jsluiceEnabled`	true / false	`true`	settable	Whether jsluice runs. It runs inside resource enumeration, on the JS files the crawlers found, so it is a silent no-op unless resource_enum is in scanModules.
`jsluiceExcludePatterns`	`free_text`	`["/_next/image","/_next/static","/_next/data","/__nextjs","/_nuxt/","/__nuxt","/`	settable	Deny-list patterns. Extensions like ".js" match the path suffix only; everything else is a substring match against the URL path and query. An empty value is replaced by the default rather than honoured.
`jsluiceExtractSecrets`	true / false	`true`	settable	Whether jsluice reports secret-shaped strings it finds in JavaScript. A feature toggle, not a credential: it decides what the tool looks for, and any secret found belongs to the target.
`jsluiceExtractUrls`	true / false	`true`	settable	Extracts URLs jsluice finds in JavaScript.
`jsluiceMaxFiles`	1–10000 count	`10000`	settable	The most results jsluice accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here. In stealth mode this is forced to `20`.
`jsluiceParallelism`	1–10 threads	`5`	settable	How many jsluice operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`jsluiceTimeout`	30–86400 seconds	`300`	settable	How long jsluice may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`jsluiceVerifyAcceptStatus`	`status_codes`	`[200,201,301,302,307,308,401,403]`	settable	HTTP status codes treated as "live" by the verifier An empty value is replaced by the default rather than honoured.
`jsluiceVerifyDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/httpx:latest`	settable	The container image spawned for jsluice. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`jsluiceVerifyRateLimit`	1–1000 rps	`50`	settable	Requests per second jsluice sends. Capped to the engagement ceiling at scan start, so a value above the ceiling is rewritten rather than refused. Capped to the engagement rate ceiling at scan start.
`jsluiceVerifyThreads`	1–500 threads	`50`	settable	How many jsluice operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`jsluiceVerifyTimeout`	1–60 seconds	`5`	settable	How long jsluice may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`jsluiceVerifyUrls`	true / false	`true`	settable	After extraction, run extracted URLs through httpx and a deny-list filter so only live, non-asset URLs reach the graph. Disable to keep all extracted URLs (legacy behavior).

Katana

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`katanaConcurrency`	1–50 threads	`15`	settable	How many katana operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`katanaCustomHeaders`	`http_header`	`["User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTM`	settable	Custom HTTP headers for crawler
`katanaDepth`	1–10 depth	`2`	settable	How many links deep the crawl goes from each seed URL. In stealth mode this is forced to `1`.
`katanaDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/katana:latest`	settable	The container image spawned for katana. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`katanaEnabled`	true / false	`true`	settable	Whether katana runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`katanaExcludePatterns`	`free_text`	`["/_next/image","/_next/static","/_next/data","/__nextjs","/_nuxt/","/__nuxt","/`	settable	Regex patterns to exclude from crawling
`katanaJsCrawl`	true / false	`true`	settable	Renders JavaScript with a headless browser, which finds routes a static crawl cannot and costs far more per page. In stealth mode this is forced to `false`.
`katanaMaxUrls`	1–1000000 count	`300000`	settable	The crawl stops after this many URLs, which is what bounds a crawl on a large site. In stealth mode this is forced to `50`.
`katanaParallelism`	1–50 threads	`8`	settable	How many katana operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`katanaParamsOnly`	true / false	`false`	settable	Only extract parameterized URLs
`katanaRateLimit`	1–100000 rps	`50`	settable	Requests per second katana sends. Capped to the engagement ceiling at scan start, so a value above the ceiling is rewritten rather than refused. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `2`.
`katanaScope`	`free_text`	`dn`	settable	Scope filter: "dn" (domain), "rdn" (root domain), etc. A Rules of Engagement document may never set this.
`katanaTimeout`	60–86400 seconds	`3600`	settable	How long katana may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

Kiterunner

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`kiterunnerBruteforceMethods`	`free_text`	`["POST","PUT","DELETE","PATCH"]`	settable	["GET", "POST", "PUT", "DELETE", "PATCH"]
`kiterunnerConnections`	1–500 threads	`100`	settable	How many kiterunner operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`kiterunnerDetectMethods`	true / false	`true`	settable	Establishes which HTTP methods each discovered route accepts, at the cost of extra requests per route.
`kiterunnerEnabled`	true / false	`false`	settable	Whether kiterunner runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`kiterunnerHeaders`	`http_header`	(empty)	settable	Add auth tokens for authenticated API scanning
`kiterunnerIgnoreStatus`	`free_text`	(empty)	settable	Blacklist: filter out noise from common errors
`kiterunnerMatchStatus`	`free_text`	`[200,201,204,301,302,401,403,405]`	settable	Whitelist: only show endpoints with these status codes (includes auth-protected)
`kiterunnerMethodDetectionMode`	`free_text`	`bruteforce`	settable	How kiterunner establishes which HTTP methods a discovered route accepts.
`kiterunnerMethodDetectRateLimit`	1–100000 rps	`50`	settable	Requests per second the kiterunner method-detection pass sends. Capped by the engagement ceiling at scan start. Capped to the engagement rate ceiling at scan start.
`kiterunnerMethodDetectThreads`	1–500 threads	`25`	settable	How many kiterunner operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`kiterunnerMethodDetectTimeout`	1–86400 seconds	`5`	settable	How long kiterunner may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`kiterunnerMinContentLength`	0–100000 count	`0`	settable	Ignore responses smaller than this (bytes)
`kiterunnerParallelism`	1–5 threads	`3`	settable	How many kiterunner operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`kiterunnerRateLimit`	1–100000 rps	`100`	settable	Requests per second kiterunner sends. Capped to the engagement ceiling at scan start, so a value above the ceiling is rewritten rather than refused. Capped to the engagement rate ceiling at scan start.
`kiterunnerScanTimeout`	60–86400 seconds	`1000`	settable	How long kiterunner may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`kiterunnerThreads`	1–500 threads	`50`	settable	How many kiterunner operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`kiterunnerTimeout`	1–86400 seconds	`10`	settable	How long kiterunner may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`kiterunnerWordlists`	`free_text`	`["routes-large"]`	settable	Wordlists: "routes-small", "routes-large"

knockpy

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`knockpyReconEnabled`	true / false	`true`	settable	Whether knockpy runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`knockpyReconMaxResults`	1–50000 count	`5000`	settable	Maximum subdomains taken from a knockpy run. In stealth mode this is lowered to at most 100; a quieter value is kept.

masscan

phase port_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`masscanAiPortCatalogEnabled`	true / false	`true`	settable	Whether masscan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`masscanBanners`	true / false	`false`	settable	Whether masscan grabs a service banner as well as recording the open port, which turns a SYN scan into a connecting one.
`masscanCustomPorts`	`free_text`	(empty)	settable	Overrides Top Ports if set. Use ranges: 8080-8090
`masscanEnabled`	true / false	`false`	settable	Whether masscan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`masscanExcludeTargets`	`free_text`	(empty)	settable	Comma-separated targets to exclude
`masscanRate`	1–100000 rps	`1000`	settable	Packets per second masscan sends. Masscan is the fastest port scanner here and the one most likely to be noticed; the engagement ceiling caps it at scan start. Capped to the engagement rate ceiling at scan start.
`masscanRetries`	0–10 count	`1`	settable	How many times masscan retries a failed request before giving up on it.
`masscanTopPorts`	`free_text`	`1000`	settable	Use “100”, “1000”, or “full” for all 65535 ports
`masscanWait`	0–86400 seconds	`10`	settable	Wait time in seconds

MITRE ATT&CK enrichment

phase vuln_scan · sends no traffic

Field	Accepts	Default	MCP	Meaning
`mitreAutoUpdateDb`	true / false	`true`	settable	Whether the MITRE dataset is refreshed before enrichment runs.
`mitreCacheTtlHours`	1–168 count	`24`	settable	Part of MITRE ATT&CK enrichment, which annotates findings with the technique they correspond to. Enrichment reads existing findings and sends no traffic.
`mitreEnabled`	true / false	`true`	settable	Whether mitre runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`mitreEnrichGvm`	true / false	`true`	settable	Attaches MITRE ATT&CK technique references to GVM findings.
`mitreEnrichRecon`	true / false	`true`	settable	Attaches MITRE ATT&CK technique references to recon findings.
`mitreIncludeCapec`	true / false	`true`	settable	Includes the CAPEC attack-pattern reference alongside each technique.
`mitreIncludeCwe`	true / false	`true`	settable	Includes the CWE weakness reference alongside each technique.

naabu

phase port_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`naabuCustomPorts`	`free_text`	(empty)	settable	Custom port list, e.g. "80,443,8080"
`naabuDisplayCdn`	true / false	`true`	settable	Reports CDN-fronted hosts rather than dropping them silently, so the operator can see what was skipped.
`naabuDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/naabu:latest`	settable	The container image spawned for naabu. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`naabuEnabled`	true / false	`true`	settable	Whether naabu runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`naabuExcludeCdn`	true / false	`false`	settable	Skips hosts that resolve to a CDN, where a port result describes the CDN rather than the target.
`naabuPassiveMode`	true / false	`false`	settable	Use Shodan InternetDB instead of active scan In stealth mode this is forced to `true`.
`naabuRateLimit`	1–100000 rps	`1000`	settable	Packets per second naabu sends, across all hosts combined. The engagement ceiling caps it at scan start, so a value above the ceiling is rewritten rather than refused. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `10`.
`naabuRetries`	0–10 count	`1`	settable	How many times naabu retries a failed request before giving up on it.
`naabuScanType`	`free_text`	`s`	settable	"s" (SYN) or "c" (connect) In stealth mode this is forced to `"c"`.
`naabuSkipHostDiscovery`	true / false	`true`	settable	Skips the ping sweep and scans every listed host directly, which is slower but finds hosts that do not answer ICMP. In stealth mode this is forced to `true`.
`naabuThreads`	1–100 threads	`25`	settable	How many naabu operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`naabuTimeout`	1000–3600000 milliseconds	`10000`	settable	How long naabu waits for a port to answer, in MILLISECONDS. At the default of 10000 a full sweep of an unresponsive host is slow; below about 1000 a distant host starts reading as closed.
`naabuTopPorts`	`free_text`	`1000`	settable	Top N ports, e.g. "1000"
`naabuVerifyPorts`	true / false	`true`	settable	Re-probes each discovered port before reporting it, which removes the false positives a fast SYN scan produces.

Netlas

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`netlasEnabled`	true / false	`false`	settable	Whether netlas runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`netlasWorkers`	1–20 threads	`5`	settable	How many netlas operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

Nmap

phase port_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`nmapAiVersionRegexEnabled`	true / false	`true`	settable	Whether nmap runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`nmapEnabled`	true / false	`true`	settable	Whether nmap runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`nmapHostTimeout`	30–86400 seconds	`300`	settable	How long nmap may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`nmapParallelism`	1–10 threads	`5`	settable	How many nmap operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.
`nmapScriptScan`	true / false	`true`	settable	Run NSE vuln scripts In stealth mode this is forced to `false`.
`nmapTimeout`	60–86400 seconds	`600`	settable	How long nmap may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`nmapTimingTemplate`	`free_text`	`T3`	settable	Nmap's timing template, T0 to T5. Higher is faster and noisier; the stealth pass forces T2. Not a wordlist despite once being classified with them. In stealth mode this is forced to `"T2"`.
`nmapVersionDetection`	true / false	`true`	settable	Detect service versions

Nuclei

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`nucleiAiResponseFilter`	true / false	`false`	settable	Use AI to classify Nuclei response bodies as WAF/rate-limit block pages when the static keyword list misses (requires aiInPipeline=true). Augments is_false_positive(); kicks in only on suspicious status codes (403/406/418/429/503) so cost stays bounded. Default false.
`nucleiAiTags`	true / false	`false`	settable	Use AI to prune the Nuclei tag list based on detected tech stack (requires aiInPipeline=true). Replaces the static nucleiTags list with a tech-aware subset chosen by the LLM at scan time. Default false.
`nucleiAutoUpdateTemplates`	true / false	`true`	settable	Whether nuclei refreshes its template set before scanning. It fetches from the public template repository, so it is an egress decision as well as a freshness one.
`nucleiBulkSize`	1–500 threads	`25`	settable	Number of hosts to process in parallel In stealth mode this is forced to `5`.
`nucleiConcurrency`	1–500 threads	`25`	settable	How many nuclei operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `2`.
`nucleiCustomTemplates`	`project_file`	(empty)	settable	Custom template paths
`nucleiDastMode`	true / false	`false`	settable	Dynamic testing mode In stealth mode this is forced to `false`.
`nucleiDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/nuclei:latest`	settable	The container image spawned for nuclei. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`nucleiEnabled`	true / false	`true`	settable	Whether nuclei runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`nucleiExcludeTags`	`free_text`	`["dos","fuzz"]`	settable	Template tags to exclude, e.g. ["dos", "fuzz"]
`nucleiExcludeTemplates`	`free_text`	(empty)	settable	Template paths to exclude
`nucleiFollowRedirects`	true / false	`true`	settable	Follows a redirect while running a template rather than matching on the redirect itself. A template that expects its payload to land on the final page needs this; one matching a redirect header does not.
`nucleiHeadless`	true / false	`false`	settable	Browser-based testing In stealth mode this is forced to `false`.
`nucleiInteractsh`	true / false	`true`	settable	Out-of-band interaction detection In stealth mode this is forced to `false`.
`nucleiMaxHostError`	1–10000 count	`30`	settable	How many errors nuclei tolerates from one host before it stops probing that host (nuclei's -mhe), so one dead host in a large target list cannot soak the whole pass in timeouts.
`nucleiMaxRedirects`	0–50 count	`10`	settable	Maximum redirect chain to follow
`nucleiMaxRuntime`	600–604800 seconds	`86400`	settable	Wall-clock ceiling for a single nuclei pass, in seconds (default 24h). A pass that reaches it is stopped, its partial findings are kept, and the run is recorded as truncated. Not a rate and has no unlimited value; raise it (up to 7 days) rather than disabling it.
`nucleiNewTemplatesOnly`	true / false	`false`	settable	Restricts the run to templates added since the last template update, for a fast re-check.
`nucleiRateLimit`	1–100000 rps	`100`	settable	Requests per second nuclei sends, across all templates and all targets combined rather than per template. Capped by the engagement ceiling at scan start. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `5`.
`nucleiRetries`	0–10 count	`1`	settable	How many times nuclei retries a failed request before giving up on it.
`nucleiScanAllIps`	true / false	`false`	settable	Scans every IP a host resolves to rather than the first, which matters behind round-robin DNS.
`nucleiSelectedCustomTemplates`	`project_file_name`	(empty)	settable	Which of this project's uploaded custom templates to run. Each entry is a path a scan container opens, so it is validated against the project's own upload directory.
`nucleiSeverity`	`info`, `low`, `medium`, `high`, `critical`, `unknown`	`["critical","high","medium","low"]`	settable	Severity filter: "critical", "high", "medium", "low"
`nucleiSystemResolvers`	true / false	`true`	settable	Uses the container's own DNS resolvers rather than nuclei's defaults.
`nucleiTags`	`free_text`	`["cve","xss","sqli","rce","lfi","ssrf","xxe","ssti"]`	settable	Template tags to include
`nucleiTagsAiUseJev`	true / false	`false`	settable	Engine for the Nuclei tag selector: false uses the LLM in aiPipelineModel, true uses TypeSafe AI Jev (model jev-1.13.0). Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one the hook uses its static fallback.
`nucleiTakeoverRunTimeout`	1–86400 seconds	`1800`	settable	How long nuclei may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`nucleiTakeoversEnabled`	true / false	`true`	settable	Whether nuclei runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`nucleiTemplates`	`free_text`	(empty)	settable	Specific template paths to include
`nucleiTimeout`	1–86400 seconds	`10`	settable	Per-request timeout, in seconds: how long nuclei waits for one template's HTTP request (nuclei's -timeout), NOT a cap on the whole run. The wall-clock ceiling for a pass is nucleiMaxRuntime.

OpenAPI ingestion

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`openapiAutoDiscover`	true / false	`false`	settable	Discover specifications from crawl results and configured paths on known in-scope base URLs. Opt-in and disabled in stealth mode. In stealth mode this is forced to `false`.
`openapiDiscoveryPaths`	`url_path_list`	`["/openapi.json","/openapi.yaml","/swagger.json","/swagger.yaml","/v3/api-docs",`	settable	Paths probed for OpenAPI documents on every in-scope origin during auto-discovery: at most 200, each starting with /, without whitespace, a query or a fragment. A path that breaks those rules is refused at the write and skipped at scan time without dropping the others. Empty disables common-path probes. A Rules of Engagement document may never set this. In stealth mode this is forced to `[]`.
`openapiEnabled`	true / false	`true`	settable	Whether to import declared API operations from OpenAPI documents. Fetches specifications but never calls declared operations. In stealth mode this is forced to `false`.
`openapiMaxDocuments`	1–200 count	`50`	settable	Maximum number of specification documents attempted per target run. In stealth mode this is lowered to at most 1; a quieter value is kept.
`openapiSources`	`json_object`	`[]`	closed	Project-specific specification sources. URLs may contain credentials; excluded from MCP reads and writes and reusable presets. Withheld from every read on the MCP surface. In stealth mode this is forced to `[]`.
`openapiTimeout`	1–60 seconds	`10`	settable	Per-request timeout in seconds when fetching a specification or same-origin reference. In stealth mode this is lowered to at most 10; a quieter value is kept.

Origin discovery

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`originDiscoveryEnabled`	true / false	`false`	settable	Whether Origin discovery runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`originDiscoveryKeyless`	true / false	`true`	settable	Non-CDN subdomains + SPF/MX + crt.sh (no key needed)
`originDiscoveryMaxCandidates`	1–200 count	`25`	settable	The most results Origin discovery accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`originDiscoveryMaxSearchCalls`	0–500 count	`50`	settable	Per-scan budget of keyed scanner-search calls
`originDiscoveryPassiveDns`	true / false	`true`	settable	SecurityTrails + ViewDNS DNS history
`originDiscoveryScanners`	true / false	`true`	settable	Favicon/cert pivots via Shodan/Censys/FOFA/ZoomEye/OTX/VirusTotal (each uses its own key) In stealth mode this is forced to `false`.
`originDiscoveryThreshold`	0–100 percent	`60`	settable	Weighted-similarity confidence (0-100) needed to confirm an origin
`originDiscoveryTimeout`	1–60 seconds	`10`	settable	How long Origin discovery may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`originDiscoveryWorkers`	1–30 threads	`10`	settable	How many Origin discovery operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

OSINT enrichment

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`osintEnrichmentEnabled`	true / false	`false`	settable	Whether OSINT enrichment runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.

ParamSpider

phase resource_enum · passive traffic

Field	Accepts	Default	MCP	Meaning
`paramspiderEnabled`	true / false	`false`	settable	Whether paramspider runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `true`.
`paramspiderPlaceholder`	`free_text`	`FUZZ`	settable	Replacement value for parameter values (e.g., FUZZ for fuzzing tools)
`paramspiderTimeout`	10–86400 seconds	`120`	settable	How long paramspider may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`paramspiderWorkers`	1–10 threads	`8`	settable	How many paramspider operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

Path traversal

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`pathTraversalArchiveExtractionEnabled`	true / false	`false`	settable	Whether Path traversal runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`pathTraversalGitReconstructEnabled`	true / false	`false`	settable	Lets the path_traversal skill reconstruct a repository from an exposed .git that recon's route exposure scan found: it walks the loose objects in the Kali sandbox (500 objects, 20 MiB, 240 s at most), scans them for secrets, reports counts and types only, and deletes them. Also needs data exfiltration allowed in the Rules of Engagement. Takes effect only while route exposure is on. Off by default.
`pathTraversalOobCallbackEnabled`	true / false	`true`	settable	Whether Path traversal runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`pathTraversalOobProvider`	`free_text`	`oast.fun`	settable	interactsh-client server. Use a self-hosted instance if oast.fun is blocked. Only used when the OOB callback workflow is enabled. A Rules of Engagement document may never set this.
`pathTraversalPayloadReferenceEnabled`	true / false	`true`	settable	Whether Path traversal runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`pathTraversalPhpWrappersEnabled`	true / false	`true`	settable	Whether Path traversal runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`pathTraversalRequestTimeout`	1–120 seconds	`10`	settable	How long Path traversal may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`pathTraversalValidateSecretsEnabled`	true / false	`false`	settable	Lets the path_traversal skill check a secret found in an exposed file with one identity-only call to its provider (AWS, GitHub, GitLab, Slack). The provider sees the call, and a canary token alerts its owner. Takes effect only while route exposure is on. Off by default.

Phishing

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`phishingSmtpConfig`	`free_text`	(empty)	settable	Free-text SMTP settings injected into the agent prompt for social engineering email delivery. The agent reads this as-is when the social engineering simulation skill is active. A Rules of Engagement document may never set this. Withheld from every read on the MCP surface.

Pipeline

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`scanModules`	`domain_discovery`, `port_scan`, `http_probe`, `resource_enum`, `vuln_scan`, `js_recon`	`["domain_discovery","port_scan","http_probe","resource_enum","vuln_scan"]`	settable	Pipeline phases to run. Values: "domain_discovery", "port_scan", "http_probe", "resource_enum", "vuln_scan". "js_recon" is still accepted so older projects and presets load, but it does nothing: JS recon runs on jsReconEnabled alone.
`stealthMode`	true / false	`false`	settable	The engagement's PASSIVE switch. On, every recon tool is forced to its passive, low-noise profile and the noisiest ones are switched off outright (apply_stealth_overrides in recon/project_settings.py, driven by the registry's own stealth profile). This is the setting a rule of engagement means by "passive only", "no active scanning", "observation only" or "do not send probes to the target": it is one switch rather than the dozens of per-tool flags such a rule would otherwise have to name.
`updateGraphDb`	true / false	`true`	closed	Whether results are stored in the graph database. Off, the scan still runs and still reaches the target, and writes nothing, so every later read reports "nothing found" where the truth is "nothing was written". That is the scanned-and-clean versus never-scanned confusion, so it is a debug switch rather than a pipeline parameter.
`useBruteforceForSubdomains`	true / false	`false`	settable	Enable DNS brute-force for subdomain discovery In stealth mode this is forced to `false`.

Pipeline AI master switch

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`aiInPipeline`	true / false	`false`	settable	Master toggle that enables AI-powered enhancements across all recon modules that support them. Cascades on/off to per-tool AI flags (ffufAiExtensions, nucleiAiTags, wafAiClassifier, nucleiAiResponseFilter, takeoverAiClassifier).
`aiPipelineModel`	`free_text`	`claude-opus-4-6`	settable	Model identifier for the AI hooks in recon (e.g. "claude-opus-4-6", "claude-haiku-4-5-20251001"). Independent of agentOpenaiModel. A Rules of Engagement document may never set this.

Port scan AI catalog

phase port_scan · sends no traffic

Field	Accepts	Default	MCP	Meaning
`portScanAiPortCatalogEnabled`	true / false	`true`	settable	Whether Port scan AI catalog runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.

Project identity

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`activationStartedAt`	`free_text`	—	closed	When the current version activation took the lock. Used to detect a stale lock left by a crashed activation.
`activationState`	`free_text`	`idle`	closed	The version-activation lock. Taken while a saved scan version rewrites the graph and released afterwards; it is a lock flag the application owns, not a setting.
`activationVersionId`	`free_text`	—	closed	The scan version currently being activated. Written by the activation state machine.
`createdAt`	`free_text`	`null`	closed	When the project row was created. Written by the database.
`createdById`	`free_text`	—	closed	The user who created the project. Populated going forward only; may be null on older rows.
`description`	`free_text`	—	settable	Free-text notes about the engagement, shown beside the project name. A Rules of Engagement document may never set this.
`id`	`free_text`	`null`	closed	The project's primary key. Row identity, not configuration.
`loadedPreset`	—	—	closed	The name of the preset last loaded into this project and a fingerprint of the settings it produced. The UI shows the preset as applied only while the saved settings still match the fingerprint. Written by the load path, not configured directly.
`mcpKaliExecEnabled`	true / false	`true`	closed	Whether MCP tokens on this project may run shell commands in the Kali sandbox. It decides what a credential can do rather than how the pipeline scans, so it is closed on the MCP surface: a token that could set it would be granting itself shell access.
`name`	`free_text`	—	settable	The project's display name. Ordinary metadata, shown in every list and header. A Rules of Engagement document may never set this.
`reconPresetId`	`free_text`	—	closed	The preset last applied to this project, recorded so the UI can show which one it came from. Written by the apply path, not configured directly.
`updatedAt`	`free_text`	—	closed	When the project row last changed. Written by the database on every update.
`updatedById`	`free_text`	—	closed	The user who last updated the project. Populated going forward only; may be null on older rows.
`userId`	`free_text`	—	closed	The owning user. Writing it reassigns the project to someone else and configures nothing, which is why it is closed on every surface. Withheld from every read on the MCP surface.

puredns

phase domain_discovery · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`purednsDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`frost19k/puredns:latest`	settable	The container image spawned for puredns. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`purednsEnabled`	true / false	`true`	settable	Whether puredns runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`purednsRateLimit`	0–100000 rps	`0`	settable	DNS queries per second puredns sends to its resolvers. ZERO MEANS UNLIMITED, so 0 is the most aggressive value available and not the safest. Under an engagement ceiling a 0 here is rewritten to the ceiling at scan start. 0 means UNLIMITED here, so it is the fastest value and not the safest. Capped to the engagement rate ceiling at scan start.
`purednsSkipValidation`	true / false	`false`	settable	Skips the wildcard-validation pass, which is much faster and lets wildcard DNS flood the results.
`purednsThreads`	0–1000 threads	`0`	settable	How many puredns operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`purednsWildcardBatch`	0–100000 count	`0`	settable	Wildcard filtering batch size

Remote code execution

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`rceAggressivePayloads`	true / false	`false`	settable	Whether the RCE technique uses payloads that can leave a running process or a changed file behind, rather than proof-of-concept ones that only echo.
`rceDeserializationEnabled`	true / false	`true`	settable	Whether rce runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`rceHeapdumpAnalysisEnabled`	true / false	`false`	settable	Lets the rce skill download an exposed Spring heap dump once (512 MiB at most), count the credential shapes in it, and delete it. Also needs data exfiltration allowed and sensitive data handling above "no access" in the Rules of Engagement. Takes effect only while route exposure is on. Off by default.
`rceOobCallbackEnabled`	true / false	`true`	settable	Whether rce runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.

Resource enum AI hints

phase resource_enum · sends no traffic

Field	Accepts	Default	MCP	Meaning
`resourceEnumAiClassifierEnabled`	true / false	`true`	settable	Whether Resource enum AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`resourceEnumAiParamInjectableFlagEnabled`	true / false	`true`	settable	Whether Resource enum AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`resourceEnumAiPathClassifierEnabled`	true / false	`true`	settable	Whether Resource enum AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`resourceEnumAiRagPathFlagEnabled`	true / false	`true`	settable	Whether Resource enum AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`resourceEnumAiToolArgPathEnabled`	true / false	`true`	settable	Whether Resource enum AI hints runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`resourceEnumJevToolHealth`	true / false	`false`	settable	Asks TypeSafe AI Jev (model jev-1.13.0) about a collection tool's empty result whose error output the deterministic checks cannot place: false asks nothing, true asks whether it describes a transient failure that a second run could fix. The deterministic checks always run and record such a result as a coverage gap whatever this says. Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account. At most 20 questions per run, with header values redacted. The verdict (transient or permanent, with its confidence) is kept in the recon output next to the coverage gap; no tool is re-run.

Route exposure scan

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`routeExposureBackupFiles`	true / false	`true`	settable	Whether the route exposure scan guesses backup copies of the server-side files recon found (name.bak, name~, .old, .orig, .save, .dist and the editor's .name.swp) plus root archives and dumps (backup.zip, backup.tar.gz, site.zip, db.sql, dump.sql, database.sql). Each guess is read with HEAD, then a 1 KB Range request, never downloaded in full; an archive, SQL dump or PHP source that differs from a made-up path and from the original file becomes a finding. At most 40 guesses per host. On by default.
`routeExposureCacheRider`	true / false	`true`	settable	Whether, on a logged-in route whose answer is personalised and sensitive, the scan checks the response's cache headers and the cache oracle for a layer that would let a logged-in page be served from cache to an anonymous visitor. It never plants or fetches from cache; that proof is the agent's. On by default.
`routeExposureCompareLoggedIn`	true / false	`true`	settable	Whether the route exposure scan also requests each route logged in, with the project's authentication profile, wherever that profile covers the host (its scope hosts, never an RoE-excluded host). The logged-in answer is sent whatever the anonymous one was: it finds routes that return too much to their own user (sensitive_data), and compares bodies to tell a personal page from a public one. Without a profile nothing is sent logged in. Logged-in requests go through the capture proxy only for bearer, basic and cookie profiles. On by default.
`routeExposureCorsRider`	true / false	`true`	settable	Whether, on a logged-in route whose answer is personalised and sensitive, the scan resends the request with crafted Origin headers (an attacker origin, null, a suffix and a prefix variant) to detect a reflected origin returned with credentials. At most 4 extra requests per route; it reads only the CORS response headers and never a body. On by default.
`routeExposureCrossHost`	true / false	`false`	settable	Whether a path that produced a candidate or finding on one in-scope host is tried on the other in-scope hosts that share its technology set or page title. Each target host must still pass the scope and SSRF checks. At most 50 requests. Off by default, because it widens the traffic across hosts.
`routeExposureEnabled`	true / false	`false`	settable	Whether the route exposure scan runs. It requests the routes recon found (crawl, ffuf, kiterunner, OpenAPI, JS bundles) plus routes it generates, anonymously and, where the authentication profile covers the host, logged in, and writes two kinds of Vulnerability under source route_exposure: leads for the agent to confirm (needs_agent_confirmation true, severity info: a route answering without credentials where authentication was expected, or a response carrying sensitive fields or other people's personal data) and findings a signature proves (exposed .git, .env, web.config, backups, debug consoles and endpoints, debug pages, route catalogs, secrets in a response). It sends GET and HEAD only and never requests a logout, delete, reset or other state- changing path. Personal data is stored as its types and one record count, never a value. It runs in GROUP 6 Phase A whenever active scanning runs (http_probe found live URLs), whatever scanModules says: there is no second level. It needs the routes resource_enum (and js_recon, when on) found. In stealth mode this is forced to `false`.
`routeExposureFollowCatalogs`	true / false	`true`	settable	Whether, when the scan finds an open route catalog (an /actuator/mappings document, an anonymously served OpenAPI or Swagger spec, /rails/info/routes), it parses the routes it lists and requests them once, in a single follow-up round that never recurses. At most 500 routes per host. On by default.
`routeExposureFrameworkSeeds`	true / false	`true`	settable	Whether the route exposure scan requests framework-specific debug and management paths based on the technologies recon fingerprinted on the host: Spring actuator endpoints (/actuator/env, /actuator/mappings, and /actuator/heapdump with HEAD only), Django /admin, Laravel /telescope, /horizon and /_ignition/health-check, Rails /rails/info/routes, Go /debug/pprof, .NET /elmah.axd, PHP /phpinfo.php, Apache /server-status. At most 30 per host. On by default.
`routeExposureGitWalk`	true / false	`true`	settable	Whether, after an exposed /.git/HEAD is confirmed, the scan fetches the ref it names and one loose commit object and inflates it with a bounded decompressor, to prove the repository is reconstructable. At most 4 requests, each body at most 64 KiB; it stores only the proof and the commit tree's hash prefix, and never downloads the tree or blobs. On by default.
`routeExposureJevAssess`	true / false	`false`	settable	Assesses each route exposure lead with TypeSafe AI Jev (model jev-1.13.0): false keeps the deterministic route class only, true asks Jev which kind of route it is and how sensitive what it returns is, from the evidence alone (the statuses, content class, masked key names and closed hints, never the route class or a value). Both answers are saved on the lead (route_jev_kind, route_jev_sensitivity) and the agent confirms the leads most sensitive first, a lead Jev reads as public content at confidence 70 or more last. It never drops a lead and never rewrites its route class. Takes effect only when aiInPipeline and routeExposureEnabled are both true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account. At most 200 leads per scan, one Jev request each, 4 per call, within a 60-second budget.
`routeExposureJevRank`	true / false	`false`	settable	Ranks the route exposure scan's routes with TypeSafe AI Jev (model jev-1.13.0): false keeps the deterministic budget order (floor first, then the prior score), true asks Jev, per route, whether it is meant only for logged-in, internal or administrative callers, and its answer fills the budget after the floor: the routes Jev reads as private are requested first, the ones it did not answer follow by prior score. The answer is saved on each finding as route_jev_privacy. It never drops a floor route and never invents one, and when Jev cannot answer the order is the deterministic one. Takes effect only when aiInPipeline and routeExposureEnabled are both true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account. At most 400 routes per scan, sent as path, method, a closed category and four flags.
`routeExposureJevWallCheck`	true / false	`false`	settable	Asks TypeSafe AI Jev (model jev-1.13.0) whether an HTML page the route exposure scan kept as a lead is really a login form, an error page or a placeholder that the deterministic page check could not place: false keeps the deterministic verdict only, true saves Jev's page class on the lead (anon_page_class) and puts a lead Jev places as a wall (a non-app class at confidence 70 or more) last in the order the agent confirms; it never drops a lead. The page body reaches Jev masked and capped. Takes effect only when aiInPipeline and routeExposureEnabled are both true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account. At most 100 pages per scan, 4 per call, within a 60-second budget.
`routeExposureMaxRoutesPerHost`	10–5000 count	`300`	settable	The most routes the route exposure scan requests on one host, per scope group of a Domain batch. Routes with a prior score of 5 or more (an OpenAPI security declaration, a discovery tool's 401/403, an internal path word) and every version-control or env guess go first; the rest are cut and counted as budget_cut.
`routeExposureMaxRoutesTotal`	50–50000 count	`3000`	settable	The most routes the route exposure scan requests in one invocation, across all hosts, per scope group of a Domain batch. Hosts take turns, so one host never spends the whole budget.
`routeExposureMaxRps`	1–50 rps	`5`	settable	Requests per second the route exposure scan sends across all hosts, shared by every worker and counting control requests, logged-in comparisons and backup reads. It runs at the same time as the other Phase A scanners (nuclei, cache poisoning, GraphQL), so the load on a target is the sum of their rates. One clock serves the whole run, so a Domain batch's groups never reset it. Capped to the engagement rate ceiling at scan start.
`routeExposurePiiDetection`	true / false	`true`	settable	Whether the route exposure scan looks for personal data in the answers it reads: e-mail addresses, phone numbers and dates of birth under their keys, card numbers passing Luhn, IBANs with a valid checksum, US SSNs and Italian fiscal codes. It stores the kinds found and the largest count of distinct values, never a value; three or more e-mails or phones in one answer mark other people's records. Sends no extra request. On by default.
`routeExposureSeedPaths`	true / false	`true`	settable	Whether the route exposure scan requests well-known internal, admin and debug paths (/internal, /api/admin/users, /actuator/env, /console, /phpinfo.php ...) at each base URL and under each API prefix recon saw (/api, /api/v1). At most 60 per host. On by default.
`routeExposureSiblings`	true / false	`true`	settable	Whether the route exposure scan generates sibling routes from each harvested API route: the collection (dropping a trailing id), its /export, /all and /list, an admin or internal segment inserted after the API prefix, and an /internal or /private prefix. At most 6 per route and 100 per host. On by default.
`routeExposureTimeout`	2–60 seconds	`10`	settable	Seconds the route exposure scan waits for one answer. A route that times out costs that route only; a host is skipped once it stops answering altogether.
`routeExposureVcsFiles`	true / false	`true`	settable	Whether the route exposure scan requests version-control and environment files at each base URL and under each top-level directory recon saw (a nested application): /.git/HEAD, /.git/config, /.svn/entries, /.svn/wc.db, /.hg/requires, /.env, /.env.local, /.env.production, /.env.bak and /web.config. At most 30 per host. These always get a request, ahead of the budget. On by default.
`routeExposureVersionVariants`	true / false	`true`	settable	Whether the route exposure scan requests the older API versions of each versioned route recon found: /api/v2/x is tried as /api/v1/x and /api/x, at most 3 per route. An older version that answers anonymously while the newer one is locked is a high-confidence lead. On by default.
`routeExposureWorkers`	1–16 threads	`4`	settable	How many routes the route exposure scan requests at once. The request rate stays capped by routeExposureMaxRps whatever this is. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.

Rules of Engagement

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`roeAllowAccountLockout`	true / false	`false`	settable	Whether a technique that can lock a real account out is permitted. Off keeps credential attacks to a single attempt per account and withholds the lockout-capable skills from the agent's options. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeAllowDataExfiltration`	true / false	`false`	closed	Whether a finding may be proven by moving data off the target. Part of the engagement RECORD: it is injected into the agent's prompt and read by a model, and no code refuses an exfiltration path on it. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeAllowDos`	true / false	`false`	settable	Whether denial-of-service techniques are permitted. Off means the DoS path is refused before it runs and the DoS skill is withheld from the agent's options, whatever their own toggles say. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeAllowPhysicalAccess`	true / false	`false`	closed	Whether the engagement covers physical access. Recorded for the report; the pipeline has no physical capability. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeAllowProductionTesting`	true / false	`true`	closed	Whether production systems are in scope. Off is a statement the operator makes about the estate, not something the scanner can verify, so it constrains what an agent may choose rather than what a tool can reach. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeAllowSocialEngineering`	true / false	`false`	settable	Whether techniques aimed at people rather than systems are permitted. Off withholds the social-engineering skills from the agent's options. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeClientContactEmail`	`free_text`	(empty)	closed	Where to reach the client contact by email. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it. Withheld from every read on the MCP surface.
`roeClientContactName`	`free_text`	(empty)	closed	The named client contact for this engagement. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it. Withheld from every read on the MCP surface.
`roeClientContactPhone`	`free_text`	(empty)	closed	Where to reach the client contact by phone. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it. Withheld from every read on the MCP surface.
`roeClientName`	`free_text`	(empty)	closed	The client the engagement is for, as it should appear in the report. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it. Withheld from every read on the MCP surface.
`roeComplianceFrameworks`	`PCI-DSS`, `HIPAA`, `SOC2`, `GDPR`, `ISO27001`	(empty)	closed	Compliance frameworks the engagement is conducted under, recorded for the report. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeCriticalFindingNotify`	true / false	`true`	closed	Whether a critical finding triggers an immediate notification rather than waiting for the report. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeDataRetentionDays`	0–3650 count	`90`	closed	How many days findings and captured data may be kept after the engagement ends. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeDocumentData`	`free_text`	—	closed	The uploaded Rules of Engagement document itself, stored so the engagement carries its own authority. A Bytes column written only by the project create and import endpoints, which receive the file. The MCP surface records the document's SHA-256 through attach_engagement_authorization instead; it never carries the bytes. Written only by POST /api/projects, POST and DELETE /api/projects/[id]/roe. Withheld from every read on the MCP surface.
`roeDocumentMimeType`	`free_text`	(empty)	closed	The media type of the uploaded Rules of Engagement document. Written only by POST /api/projects, POST and DELETE /api/projects/[id]/roe. Withheld from every read on the MCP surface.
`roeDocumentName`	`free_text`	(empty)	closed	The filename of the uploaded Rules of Engagement document. Written only by POST /api/projects, POST and DELETE /api/projects/[id]/roe. Withheld from every read on the MCP surface.
`roeEmergencyContact`	`free_text`	(empty)	closed	Who to reach out of hours if a scan causes an incident. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it. Withheld from every read on the MCP surface.
`roeEnabled`	true / false	`false`	closed	DERIVED, never written. Engagement limits apply when there is a limit to apply: a non-zero roeGlobalMaxRps, OR a non-empty roeExcludedHosts, OR roeTimeWindowEnabled. The column is kept so old rows and old exports still load, and every service computes the value through recon_settings.engagement.derive_roe_enabled instead of reading it. A writable master switch would be a bypass shipped as a checkbox: one write of false and the ceiling, the exclusions and the window all stop applying at once while every other field still shows its configured value. An engagement LIMIT: enforced at scan start whatever tuning says. DERIVED, never written.
`roeEngagementEndDate`	`free_text`	(empty)	closed	The last date the engagement authorises any activity. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeEngagementStartDate`	`free_text`	(empty)	closed	The first date the engagement authorises any activity. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeEngagementType`	`external`, `internal`, `web_app`, `api`, `mobile`, `physical`, `social_engineering`, `red_team`	`external`	closed	How the engagement is classified in the report: pentest, bug bounty, red team and so on. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeExcludedHostReasons`	`free_text`	(empty)	settable	Why each excluded host is excluded, parallel to roeExcludedHosts. Read by the report; it does not itself exclude anything. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeExcludedHosts`	`free_text`	(empty)	settable	Hosts that must never be touched even though they fall inside the target scope. Each entry is an exact host or IP; a domain, which also excludes its subdomains; *.domain, which excludes its subdomains only; or a CIDR. Case, port, scheme, path and a trailing dot are ignored. Enforced in three places in recon/main.py: expanded IPs, discovered subdomains and the target domain itself. A non-empty list is one of the three things that make the engagement limits live at all. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeForbiddenCategories`	`brute_force`, `dos`, `social_engineering`, `physical`, `exploitation`	(empty)	settable	Technique categories the agent may not run on this engagement. Each token is expanded to the tools that perform it and refused before the tool executes, on every executor path (CATEGORY_TOOL_MAP in agentic/orchestrator_helpers/roe_gate.py), and the list is also withheld from the classifier's options. What each token covers: brute_force is any credential guessing, including brute forcing, password spraying and credential stuffing; dos is anything aimed at availability, including denial of service, resource exhaustion and volumetric or high-volume testing; social_engineering is anything aimed at a person rather than a system, including phishing, vishing, smishing and pretexting; physical is physical premises, hardware or tailgating; exploitation is taking a confirmed finding further than proving it. The vocabulary is closed because the gate matches these tokens EXACTLY: a free-text category such as "Denial of Service" reaches the agent's prompt as advice and never reaches the gate, so the refusal the UI promises silently does not happen. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeForbiddenTools`	`cve_intel`, `execute_amass`, `execute_arjun`, `execute_code`, `execute_curl`, `execute_ffuf`, `execute_gau`, `execute_guarddog`, `execute_httpx`, `execute_hydra`, `execute_jsluice`, `execute_katana`, `execute_naabu`, `execute_nmap`, `execute_nuclei`, `execute_osv_scanner`, `execute_playwright`, `execute_subfinder`, `execute_wpscan`, `google_dork`, `kali_shell`, `metasploit_console`, `msf_restart`, `proxy_brain`, `query_graph`, `shodan`, `tradecraft_lookup`, `web_search`	(empty)	settable	Tools the agent may not run on this engagement, by name, whatever their own enable flag says. Refused before the tool executes, on every executor path, by agentic/orchestrator_helpers/roe_gate.py. The vocabulary is closed to the agent's real tool names because that gate is an exact string match: a near-miss like "hydra" for execute_hydra never matches, and the tool runs with the ban recorded and unenforced. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeGlobalMaxRps`	0–10000 count	`0`	settable	The engagement's request-rate ceiling, in requests per second. Every per-tool rate is capped to it at scan start, including the ones whose own value means 'unlimited'. ZERO MEANS NO CEILING, so this at 0 is not a slow engagement, it is an unlimited one. A non-zero value is also one of the three things that make the engagement limits live at all. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeIncidentProcedure`	`free_text`	(empty)	closed	What to do if the engagement causes an outage or a security incident. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeMaxSeverityPhase`	`informational`, `exploitation`, `post_exploitation`	`post_exploitation`	settable	The most severe phase the engagement permits, which bounds how far an attack chain may be taken: informational, exploitation or post_exploitation. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeNotes`	`free_text`	(empty)	closed	Free-text engagement notes that did not fit another field. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeParsedJson`	`json_object`	—	closed	The structured form extracted from the uploaded document, which populated the fields above. Written only by POST /api/projects, POST and DELETE /api/projects/[id]/roe. Withheld from every read on the MCP surface.
`roeRawText`	`free_text`	(empty)	closed	The Rules of Engagement document as plain text, kept so the report can quote it. Written only by POST /api/projects, POST and DELETE /api/projects/[id]/roe. Withheld from every read on the MCP surface.
`roeRequireDataEncryption`	true / false	`true`	closed	Whether engagement data must be encrypted at rest. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeSensitiveDataHandling`	`no_access`, `prove_access_only`, `limited_collection`, `full_access`	`no_access`	closed	The agreed handling rule for sensitive data encountered during the engagement. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeStatusUpdateFrequency`	`daily`, `weekly`, `on_finding`, `none`	`daily`	closed	How often the client expects a progress update during the engagement. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeThirdPartyProviders`	`free_text`	(empty)	closed	Third-party providers hosting in-scope assets, who may need their own notification. Part of the engagement RECORD. A person writes it, a model reads it, nothing enforces it; the MCP surface never touches it.
`roeTimeWindowDays`	`free_text`	`["monday","tuesday","wednesday","thursday","friday"]`	settable	Days of the week on which scanning is permitted, lower-case English names. Empty means every day, once the window is switched on. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeTimeWindowEnabled`	true / false	`false`	settable	Whether scanning is restricted to an agreed time window. The recon pipeline and the orchestrator both enforce it, and the orchestrator refuses a scan start outside the window with HTTP 403 rather than queueing it. Switching it on is one of the three things that make the engagement limits live at all. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeTimeWindowEndTime`	`free_text`	`18:00`	settable	Local end of the permitted scanning window, as HH:MM. An end before the start is read as a window that crosses midnight. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeTimeWindowStartTime`	`free_text`	`09:00`	settable	Local start of the permitted scanning window, as HH:MM. An engagement LIMIT: enforced at scan start whatever tuning says.
`roeTimeWindowTimezone`	`free_text`	`UTC`	settable	The IANA timezone the scanning window's times are read in. An engagement LIMIT: enforced at scan start whatever tuning says.

Security checks

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`securityCheckAdminPortExposed`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckBasicAuthNoTls`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckCacheControlMissing`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckCspUnsafeInline`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckDatabaseExposed`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckDirectIpHttp`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckDirectIpHttps`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckDmarcMissing`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckDnssecMissing`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckEnabled`	true / false	`true`	settable	Whether Security checks runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`securityCheckInsecureFormAction`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckIpApiExposed`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckKubernetesApiExposed`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckLoginNoHttps`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckMaxWorkers`	1–50 threads	`10`	settable	How many security checks run concurrently.
`securityCheckMissingCoep`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckMissingCoop`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckMissingCorp`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckMissingPermissionsPolicy`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckMissingReferrerPolicy`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckNoRateLimiting`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckRedisNoAuth`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckSessionNoHttponly`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckSessionNoSecure`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckSmtpOpenRelay`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckSpfMissing`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTimeout`	1–86400 seconds	`10`	settable	How long Security checks may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`securityCheckTlsExpired`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTlsExpiringSoon`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTlsExpiryDays`	1–365 count	`30`	settable	How many days before expiry a certificate starts being reported as expiring soon.
`securityCheckTlsHostnameMismatch`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTlsSelfSigned`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTlsWeakCipher`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTlsWeakVersion`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckTlsWildcardOverbroad`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest.
`securityCheckWafBypass`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.
`securityCheckZoneTransfer`	true / false	`true`	settable	Whether this individual check runs during the security-check pass. Each check is one condition tested against what the earlier phases already found; turning it off removes that finding class without affecting the rest. In stealth mode this is forced to `false`.

Serialized object scan

phase js_recon · passive traffic

Field	Accepts	Default	MCP	Meaning
`serializedScanEnabled`	true / false	`false`	settable	Whether the serialized-object scan runs. It is passive and in-memory: it scans the response headers, Set-Cookie, enumerated parameters and crawled form fields the pipeline already holds for serialized-object signatures (Java, PHP, Python, .NET, Ruby, Hessian, polymorphic JSON/XML/YAML), never deserializes, and writes one info-severity :Vulnerability candidate per sink and format for the agent's deserialization skill to confirm out of band.
`serializedScanJevRank`	true / false	`false`	settable	Assesses the serialized-object candidates with TypeSafe AI Jev (model jev-1.13.0): false leaves them as the deterministic signatures flagged them, true asks Jev, per distinct blob, which serialization format it is and how likely it is to be an attacker-reachable sink, writes both onto the candidate (deser_jev_format, deser_jev_exploitability) and orders the candidates so the agent's deserialization skill confirms the most reachable first. Jev sees the blob's evidence only, never the format the signatures matched. It never drops a candidate and never changes the format the signatures matched. Takes effect only when aiInPipeline and serializedScanEnabled are both true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one, or for a blob Jev cannot answer, the candidates stand exactly as the signatures flagged them. No traffic to the target: at most 200 distinct blobs per scan, request-side ones first, within a 60-second budget checked between calls.

Server-side request forgery

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`ssrfCloudMetadataEnabled`	true / false	`true`	settable	Whether ssrf runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`ssrfCloudProviders`	`free_text`	`aws,gcp,azure,digitalocean,alibaba`	settable	Comma-separated cloud providers to include in the metadata-pivots section. Filters which provider endpoint tables ship in the prompt. Ignored when cloud metadata is disabled.
`ssrfCustomInternalTargets`	`free_text`	(empty)	settable	One hostname or IP[:port] per line. Injected into the prompt so the agent prioritizes these alongside the generic loopback / RFC1918 sweep. A Rules of Engagement document may never set this.
`ssrfDnsRebindingEnabled`	true / false	`true`	settable	Whether ssrf runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`ssrfGopherEnabled`	true / false	`true`	settable	Whether ssrf runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`ssrfInternalRanges`	`free_text`	`127.0.0.0/8,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16,169.254.0.0/16`	settable	Comma-separated CIDR blocks the agent treats as internal. Adjust for orgs using non-standard internal addressing (e.g., 100.64.0.0/10 carrier-grade NAT). A Rules of Engagement document may never set this.
`ssrfOobCallbackEnabled`	true / false	`true`	settable	Whether ssrf runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`ssrfOobProvider`	`free_text`	`oast.fun`	settable	interactsh-client server. Use a self-hosted instance if oast.fun is blocked. A Rules of Engagement document may never set this.
`ssrfPayloadReferenceEnabled`	true / false	`true`	settable	Whether ssrf runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`ssrfPortScanPorts`	`free_text`	`22,80,443,2375,3306,5432,6379,8080,8500,9200,27017`	settable	Comma-separated ports to probe via SSRF. Trim for quieter scans, extend for thorough coverage.
`ssrfRequestTimeout`	1–120 seconds	`10`	settable	How long ssrf may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

Shodan

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`shodanDomainDns`	true / false	`false`	settable	Pulls Shodan's own DNS record set for the domain.
`shodanEnabled`	true / false	`true`	settable	Whether shodan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`shodanHostLookup`	true / false	`true`	settable	Looks each discovered IP up in Shodan, which costs one API credit per host.
`shodanPassiveCves`	true / false	`true`	settable	Passive CVE lookup via Shodan
`shodanReverseDns`	true / false	`true`	settable	Asks Shodan for the names an IP is known by, which often reveals hosts DNS enumeration missed.
`shodanWorkers`	1–20 threads	`5`	settable	How many shodan operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

SQL injection

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`sqliLevel`	1–5 count	`1`	settable	Higher levels test more injection points (headers, cookies). Default: 1.
`sqliRisk`	1–3 count	`1`	settable	Higher risk uses more aggressive payloads (e.g., OR-based). Default: 1.
`sqliTamperScripts`	`free_text`	(empty)	settable	Comma-separated SQLMap tamper scripts for WAF bypass. Leave empty for auto-detection.

Subdomain discovery

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`subdomainDiscoveryEnabled`	true / false	`true`	settable	Whether Subdomain discovery runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In a Domain batch it is narrower still: it governs only the groups whose hostname carried a "*" marker, because every other group scans the exact hostnames it was given whatever this flag says.

Subdomain takeover

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`subdomainTakeoverEnabled`	true / false	`false`	settable	Whether takeover runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`takeoverAiClassifier`	true / false	`false`	settable	Use AI to disambiguate takeover findings from WAF "no host" block pages that match the same static fingerprint (requires aiInPipeline=true). For each finding the scanner probes the host, short-circuits on third-party vendor tokens, and otherwise asks the LLM to classify the body. AI-flagged collisions get score -40 (lands in manual_review). Default false.
`takeoverAiUseJev`	true / false	`false`	settable	Engine for the takeover disambiguator: false uses the LLM in aiPipelineModel, true uses TypeSafe AI Jev (model jev-1.13.0). Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one the hook uses its static fallback.
`takeoverCertValidationEnabled`	true / false	`true`	settable	Whether takeover runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`takeoverConfidenceThreshold`	0–100 percent	`60`	settable	0..100. Findings >= threshold+10 are confirmed, >= threshold are likely, otherwise manual_review
`takeoverManualReviewAutoPublish`	true / false	`false`	settable	Publish manual_review findings into the main Vulnerability stream (default false)
`takeoverRateLimit`	1–500 rps	`50`	settable	Requests per second takeover sends. Capped to the engagement ceiling at scan start, so a value above the ceiling is rewritten rather than refused. Capped to the engagement rate ceiling at scan start. In stealth mode this is forced to `10`.
`takeoverSeverity`	`info`, `low`, `medium`, `high`, `critical`, `unknown`	`["critical","high","medium"]`	settable	Severity filter for Nuclei takeover templates (e.g. ["critical","high","medium"])

subfinder

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`subfinderDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/subfinder:latest`	settable	The container image spawned for subfinder. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`subfinderEnabled`	true / false	`true`	settable	Whether subfinder runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`subfinderMaxResults`	1–50000 count	`5000`	settable	Maximum subdomains taken from a subfinder run. In stealth mode this is lowered to at most 100; a quieter value is kept.

subjack

phase vuln_scan · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`subjackAll`	true / false	`false`	settable	Test every URL, not just identified CNAMEs - slower (default false) In stealth mode this is forced to `false`.
`subjackCheckAr`	true / false	`false`	settable	Detect stale A records pointing to dead cloud IPs (manual review)
`subjackCheckMail`	true / false	`false`	settable	Check SPF include + MX takeovers In stealth mode this is forced to `true`.
`subjackCheckNs`	true / false	`false`	settable	Detect NS takeovers (expired nameservers / dangling cloud DNS delegations) In stealth mode this is forced to `true`.
`subjackEnabled`	true / false	`true`	settable	Whether subjack runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`subjackRunTimeout`	1–86400 seconds	`900`	settable	How long subjack may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`subjackSsl`	true / false	`true`	settable	Force HTTPS probing for higher accuracy (default true)
`subjackThreads`	1–100 threads	`10`	settable	How many subjack operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `3`.
`subjackTimeout`	1–86400 seconds	`30`	settable	How long subjack may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.

Supply chain recon

standalone · passive traffic

Field	Accepts	Default	MCP	Meaning
`scaIntelCorrelationEnabled`	true / false	`true`	create-only	Whether supply-chain findings are correlated against the wider intelligence set. It reaches beyond this project's own assets, so it moves with the scope. Fixed at project creation; refused on an existing project.
`supplyChainReconDeepAnalysisEnabled`	true / false	`false`	settable	Whether Supply chain recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`supplyChainReconEcosystems`	`free_text`	`npm`	settable	Comma-separated OSV ecosystems to report (default "npm"). Matched EXACTLY against the harvested package ecosystem, so use these names verbatim, case included: npm, PyPI, Go, Maven, crates.io, Packagist, RubyGems, NuGet. Empty string means no filter (report every ecosystem)
`supplyChainReconEnabled`	true / false	`false`	settable	Whether Supply chain recon runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.

Supply chain scan

standalone · passive traffic

Field	Accepts	Default	MCP	Meaning
`supplyChainDeepAnalysisEnabled`	true / false	`false`	settable	Whether Supply chain scan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`supplyChainEcosystems`	`free_text`	`npm,PyPI,Go,Maven,crates.io,Packagist,RubyGems,NuGet`	settable	Package ecosystems the supply-chain scan analyses, such as npm, pypi or go.
`supplyChainInputMode`	`upload`, `github`, `org`	`upload`	settable	Whether the supply-chain scan reads an uploaded SBOM, a repository, or an organisation.
`supplyChainOrgDeepAnalysisEnabled`	true / false	`false`	settable	Whether Supply chain scan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`supplyChainOrgIncludeArchived`	true / false	`false`	settable	Whether archived repositories are included when scanning an organisation.
`supplyChainOrgIncludeForks`	true / false	`false`	settable	Whether forked repositories are included when scanning an organisation.
`supplyChainOrgMaxRepos`	0–10000 count	`50`	settable	Maximum repositories walked when scanning an organisation.
`supplyChainOrgName`	`free_text`	(empty)	create-only	The GitHub organisation whose repositories the supply-chain scan walks. Points the scanner at someone else's assets, so after creation only the project form or update_project_scope changes it. Fixed at project creation; refused on an existing project.
`supplyChainOrgRef`	`free_text`	(empty)	settable	The git ref checked out in each organisation repository, when one is pinned.
`supplyChainRepoRef`	`free_text`	(empty)	create-only	The git ref checked out in the scanned repository. Changed together with the repository it belongs to. Fixed at project creation; refused on an existing project.
`supplyChainRepoScope`	`free_text`	(empty)	create-only	Which parts of the repository the scan reads. Changed together with the repository it belongs to. Fixed at project creation; refused on an existing project.
`supplyChainRepoUrl`	`free_text`	(empty)	create-only	The repository the supply-chain scan clones. It becomes a git clone argument against a third party's infrastructure, so like the scope it is changed after creation only by the project form or by update_project_scope. Fixed at project creation; refused on an existing project.
`supplyChainSbomFile`	`free_text`	(empty)	closed	The uploaded SBOM this project analyses. Written only by the upload endpoint, which also places the file on disk; a second writer could name a file the project never uploaded. Written only by /api/supply-chain/[projectId]/upload.
`supplyChainTyposquatEnabled`	true / false	`false`	settable	Whether Supply chain scan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.

Targeting and scope

standalone · sends no traffic

Field	Accepts	Default	MCP	Meaning
`domainBatchGroups`	`json_object`	—	closed	The legacy record of the batch scope, [{ rootDomain, prefixes[], hosts[], wildcard }], recomputed server-side from domainBatchHosts at every save and never trusted from a body. domainBatchHosts is the source of truth; the scope every service reads is scopeGroups, which GET /api/projects/[id] derives from it on each read. Kept for readers that have not moved and for a rollback. A prefix is a subdomain label, or one of two sentinels: "." meaning the scan root itself, and "*" meaning enumerate it.
`domainBatchHosts`	`free_text`	(empty)	create-only	Raw hostname list as entered or uploaded. What the operator edits. An entry may carry a wildcard marker - ".example.com" or "example.com" - which makes that one domain be fully enumerated instead of scanned as listed; every other entry stays literal. Editable after creation from the project form, and over MCP only by update_project_scope under its own permission: update_recon_settings and a parsed Rules of Engagement document can never re-point the platform's targets. Fixed at project creation; refused on an existing project.
`domainBatchMode`	true / false	`false`	create-only	Domain batch: the third targeting mode. Mutually exclusive with ipMode. One full recon walks the derived groups in order, appending to one graph. Fixed at project creation; refused on an existing project.
`ipMode`	true / false	`false`	create-only	Switches targeting from a domain to the explicit IP list. Mutually exclusive with domainBatchMode, and part of the scope, so it is write-once. Fixed at project creation; refused on an existing project.
`ownershipToken`	`free_text`	`your-secret-token-here`	create-only	The secret published in the ownership TXT record. Compared against what DNS returns, so it proves the scope rather than configuring a tool. Fixed at project creation; refused on an existing project. Withheld from every read on the MCP surface.
`ownershipTxtPrefix`	`free_text`	`_redamon-verify`	create-only	The label the ownership TXT record is looked up under, prepended to the target domain. Moves with the scope. Fixed at project creation; refused on an existing project.
`subdomainList`	`free_text`	(empty)	create-only	Hosts seeded into the scan in addition to whatever discovery finds. Part of the engagement scope, so it is write-once. Empty entries are dropped and every entry is whitespace-stripped at load. Fixed at project creation; refused on an existing project.
`targetDomain`	`free_text`	(empty)	create-only	The single root domain this engagement covers. THE scope column: every phase derives its hosts from it and the target guardrail measures against it. Set once at creation and immutable afterwards, because changing it on an existing project turns a rescan credential into one aimed at a third party. Leading and trailing whitespace is stripped at load. Fixed at project creation; refused on an existing project.
`targetGuardrailEnabled`	true / false	`true`	create-only	The hard guardrail that refuses to scan a host outside the declared scope. Turning it off removes the control that keeps the pipeline inside its engagement, which is why it moves with the scope and cannot change on an existing project. Fixed at project creation; refused on an existing project.
`targetIps`	`free_text`	(empty)	create-only	The IP addresses or CIDR ranges scanned when ipMode is on. Part of the engagement scope, so it is write-once. Empty entries are dropped and every entry is stripped at load. Fixed at project creation; refused on an existing project.
`verifyDomainOwnership`	true / false	`false`	create-only	Requires a DNS TXT record proving control of the target before any active phase runs. Part of the scope's proof, so it is write-once. Fixed at project creation; refused on an existing project.

tlsx

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`tlsxCipherConcurrency`	1–500 threads	`10`	settable	How many tlsx operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`tlsxCipherEnum`	true / false	`false`	settable	Enumerate weak supported ciphers (extra connections per target). Default false. In stealth mode this is forced to `false`.
`tlsxConcurrency`	1–300 threads	`50`	settable	How many tlsx operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `5`.
`tlsxDelay`	`free_text`	(empty)	settable	Optional -delay between connections (e.g. "200ms"); stealth only. Default "".
`tlsxDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/tlsx:latest`	settable	The container image spawned for tlsx. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`tlsxEnabled`	true / false	`true`	settable	Whether tlsx runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`tlsxIncludeHttpPorts`	true / false	`false`	settable	Also grab certs on HTTP/HTTPS ports httpx already covers (duplicate handshakes). Default false.
`tlsxInjectHostnames`	true / false	`true`	settable	Merge in-scope certificate SAN hostnames into dns.subdomains as new probe targets. Default true.
`tlsxMaxHostnamesPerIp`	0–1000 count	`1`	settable	When one IP serves several known hostnames, how many to scan (each may present a different cert). Default 1.
`tlsxMaxInjectedHostnames`	0–10000 count	`200`	settable	Cap on in-scope hostnames merged back into dns.subdomains as scan targets from certificate SANs and from JavaScript recon, applied to each source separately. Default 200.
`tlsxMaxTargets`	1–100000 count	`2000`	settable	Hard cap on ip:port pairs tlsx scans. Default 2000.
`tlsxProbeJarm`	true / false	`false`	settable	Compute JARM + JA3 fingerprints (~10 extra handshakes per target, loud). Default false. In stealth mode this is forced to `false`.
`tlsxRetries`	0–10 count	`1`	settable	How many times tlsx retries a failed request before giving up on it.
`tlsxRevPtrSni`	true / false	`false`	settable	Reverse-PTR to derive an SNI for bare IPs (extra DNS lookups). Default false.
`tlsxRunTimeout`	1–86400 seconds	`900`	settable	How long tlsx may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`tlsxScanMode`	`free_text`	`auto`	settable	TLS stack: ctls\|ztls\|openssl\|auto. Default auto.
`tlsxTimeout`	1–60 seconds	`5`	settable	How long tlsx may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`tlsxVersionEnum`	true / false	`false`	settable	Enumerate every supported TLS version (extra connections per target). Default false. In stealth mode this is forced to `false`.

TruffleHog

standalone · passive traffic

Field	Accepts	Default	MCP	Meaning
`trufflehogAllowVerificationOverlap`	true / false	`false`	settable	Whether several detectors may verify the same candidate, which costs requests but catches more.
`trufflehogArchiveMaxDepth`	0–50 depth	`0`	settable	How many nested archives are extracted before the scan stops descending.
`trufflehogArchiveMaxSize`	`free_text`	(empty)	settable	The largest archive extracted, in bytes. A larger one is skipped rather than exhausting the container.
`trufflehogArchiveTimeout`	`free_text`	(empty)	settable	Seconds one archive extraction may take before it is abandoned.
`trufflehogConcurrency`	1–500 threads	`8`	settable	How many sources trufflehog scans at once.
`trufflehogDetectorTimeout`	`free_text`	(empty)	settable	Seconds one detector may spend verifying a candidate before it is abandoned.
`trufflehogDropUnverifiedJwt`	true / false	`false`	settable	Discards JWTs that could not be verified, which are the bulk of the false positives in this class.
`trufflehogEnabled`	true / false	`false`	settable	Whether trufflehog runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`trufflehogExcludeDetectors`	`free_text`	(empty)	settable	Detectors that never run, whatever the include list says.
`trufflehogFilterEntropy`	`free_text`	(empty)	settable	The Shannon entropy below which a candidate string is discarded before detection runs.
`trufflehogForceSkipArchives`	true / false	`false`	settable	Skips archives entirely rather than extracting them, which is faster and misses anything inside.
`trufflehogForceSkipBinaries`	true / false	`false`	settable	Skips binary files rather than scanning them for embedded strings.
`trufflehogIncludeDetectors`	`free_text`	(empty)	settable	Restricts the run to these detectors. Empty means every detector runs.
`trufflehogMaxDecodeDepth`	0–50 depth	`5`	settable	How many nested encodings are unwrapped before giving up on a candidate.
`trufflehogNoVerification`	true / false	`false`	settable	Skip live verification entirely. When set, every finding lands `unverified` (never checked), which is NOT the same as "not live".
`trufflehogResultTypes`	`free_text`	`verified,unverified,unknown`	settable	Which result statuses to report: a comma-separated subset of verified,unverified,unknown,filtered_unverified. Replaces the old trufflehogOnlyVerified boolean, which combined with trufflehogNoVerification asked TruffleHog to report only verified results while forbidding it to verify — zero findings, no error.

uncover

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`uncoverDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`projectdiscovery/uncover:latest`	settable	The container image spawned for uncover. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`uncoverEnabled`	true / false	`false`	settable	Whether uncover runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`uncoverMaxResults`	1–50000 count	`50000`	settable	The most results uncover accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here. An empty value is replaced by the default rather than honoured.

urlscan.io

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`urlscanEnabled`	true / false	`true`	settable	Whether urlscan runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`urlscanMaxResults`	1–50000 count	`50000`	settable	The most results urlscan accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here. In stealth mode this is lowered to at most 100; a quieter value is kept.

Vhost / SNI enumeration

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`vhostSniBaselineSizeTolerance`	0–10000 bytes	`50`	settable	Body size deltas within this many bytes are not flagged (suppresses Set-Cookie / timestamp jitter). Default 50.
`vhostSniConcurrency`	1–100 threads	`20`	settable	How many Vhost / SNI enumeration operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`vhostSniCustomWordlist`	`hostname_list`	(empty)	settable	Optional custom candidates as inline text, one per line (not a file path); lines starting with # are ignored. Bare prefixes expand as {prefix}.{target_apex}; full hostnames are used as-is. Default "".
`vhostSniEnabled`	true / false	`false`	settable	Whether Vhost / SNI enumeration runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`vhostSniInjectDiscovered`	true / false	`true`	settable	When a hidden vhost is confirmed, create a BaseURL node so a follow-up partial recon (Katana, Nuclei) can scan it. Default true.
`vhostSniMaxCandidatesPerIp`	10–50000 count	`2000`	settable	Hard cap on candidates per IP to bound run time. Default 2000.
`vhostSniTestL4`	true / false	`true`	settable	Run the TLS SNI trick via curl --resolve (catches modern reverse proxies, k8s ingress, Cloudflare). Default true.
`vhostSniTestL7`	true / false	`true`	settable	Run the HTTP Host header trick (catches classic Apache/Nginx vhosts). Default true.
`vhostSniTimeout`	1–30 seconds	`3`	settable	How long Vhost / SNI enumeration may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`vhostSniUseDefaultWordlist`	true / false	`true`	settable	Use the bundled vhost-common.txt wordlist (~2,300 admin/dev/staging/internal/modern-stack prefixes, expanded as {prefix}.{target_apex}). Default true.
`vhostSniUseGraphCandidates`	true / false	`true`	settable	Pull hostnames from existing Subdomain, ExternalDomain, TLS SAN list, CNAME targets and reverse-DNS PTR records resolving to each target IP. Highest signal source. Default true.

VirusTotal

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`virusTotalEnabled`	true / false	`false`	settable	Whether virustotal runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`virusTotalWorkers`	1–10 threads	`3`	settable	How many virustotal operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.

WAF detection

phase http_probe · sends no traffic

Field	Accepts	Default	MCP	Meaning
`wafAiClassifier`	true / false	`false`	settable	Use AI to classify WAF/CDN presence from response headers/body/latency when the static check misses (requires aiInPipeline=true). Augments _has_cdn_markers and check_waf_bypass; AI-detected bypasses are tagged detection_method=ai_classifier (jev_classifier when wafAiUseJev is on). Default false.
`wafAiUseJev`	true / false	`false`	settable	Engine for the WAF classifier: false uses the LLM in aiPipelineModel, true uses TypeSafe AI Jev (model jev-1.13.0). Takes effect only when aiInPipeline is true, and aiInPipeline does not reset it. Needs a Jev token on the project owner's account; without one the hook uses its static fallback.

Wappalyzer

phase http_probe · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`wappalyzerAutoUpdate`	true / false	`true`	settable	Wappalyzer package version for tech database
`wappalyzerCacheTtlHours`	1–8760 count	`24`	settable	How long to cache tech database (0 = always fresh)
`wappalyzerEnabled`	true / false	`true`	settable	Whether wappalyzer runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`wappalyzerMinConfidence`	0–100 percent	`50`	settable	The confidence percentage below which a Wappalyzer technology match is discarded.
`wappalyzerNpmVersion`	`free_text`	`6.10.56`	settable	The npm version of Wappalyzer the scan container installs at run time. It is fetched from the public registry during the scan, so it decides what code runs inside the container. A Rules of Engagement document may never set this.
`wappalyzerRequireHtml`	true / false	`true`	settable	Whether a technology match requires an HTML response, which suppresses matches on JSON and binary endpoints.

Web cache poisoning

standalone · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`webCachePoisonAllowCpdos`	true / false	`false`	settable	Cache-poisoned DoS tests (oversized/meta-char). Only honored in research profile. Default false. In stealth mode this is forced to `false`.
`webCachePoisonAllowDeception`	true / false	`true`	settable	Web cache deception (path-confusion .css tricks). Default true.
`webCachePoisonAllowFrameworkPacks`	true / false	`true`	settable	Fire Next.js/Nuxt/Remix-specific vectors only when the tech fingerprint matches. Default true.
`webCachePoisonBehavioralDelay`	1–10 seconds	`1.1`	settable	Seconds to wait between the two frozen-Date probes. Larger = more reliable on slow origins, but adds latency per silent URL. Default 1.1.
`webCachePoisonBehavioralOracle`	true / false	`true`	settable	Detect silent caches (no cache-status headers, common with default Varnish/nginx or hardened CDNs) via a frozen-Date probe: fetch twice across a short delay; a frozen Date header + identical body implies a cache. Catches caches header detection alone would miss. Default true.
`webCachePoisonCacheBusterParam`	`free_text`	`rdmncb`	settable	Query param name used to isolate every test into its own cache slot. Default "rdmncb".
`webCachePoisonCacheHeader`	`free_text`	(empty)	settable	Override the cache hit/miss header if the target uses a non-standard one. Default "".
`webCachePoisonConcurrency`	1–50 threads	`10`	settable	How many Web cache poisoning operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`webCachePoisonConfirmWorkers`	1–16 threads	`6`	settable	How many Web cache poisoning operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`webCachePoisonCrossVantage`	true / false	`false`	settable	Re-confirm from a second network vantage (requires extra egress infra). Default false.
`webCachePoisonDifferential`	true / false	`true`	settable	Non-reflective detection. Beyond looking for a benign canary echoed back, also flag a poison that changes the response behaviour (a persisted status code / Location redirect / body) with no echoed marker, guarded against false positives by requiring the affected response dimension to be stable across two clean baselines. Catches CPDoS-style status flips, scheme/redirect poisoning, and cache-key confusion. Adds one extra baseline request per vector. Default true.
`webCachePoisonDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`redamon-wcvs:latest`	settable	The container image spawned for Web cache poisoning. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`webCachePoisonEnabled`	true / false	`false`	settable	Whether Web cache poisoning runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`webCachePoisonMaxRpsPerHost`	0–1000 rps	`0`	settable	Requests per second the web-cache-poisoning scan sends to any ONE host. ZERO MEANS UNLIMITED, so 0 is the most aggressive value available and not the safest. Under an engagement ceiling a 0 here is rewritten to the ceiling at scan start. 0 means UNLIMITED here, so it is the fastest value and not the safest. Capped to the engagement rate ceiling at scan start.
`webCachePoisonMinConfidence`	0–1 ratio	`0.8`	settable	Only findings scoring at or above this (0-1) become Vulnerability nodes. 0.8 keeps Confirmed + Strong. Default 0.8.
`webCachePoisonScanProfile`	`free_text`	`safe-confirm`	settable	"safe-confirm" (production, benign canaries, always isolated), "extended" (owned test targets), or "research" (lab only, enables CPDoS if allowed). Default "safe-confirm".
`webCachePoisonTimeout`	1–86400 seconds	`1800`	settable	How long Web cache poisoning may spend before it is abandoned, in seconds. A run that hits it contributes whatever it found so far rather than failing the scan.
`webCachePoisonTimeoutPerReq`	1–86400 seconds	`10`	settable	Native confirmation per-request timeout in seconds. Default 10.
`webCachePoisonVerifySsl`	true / false	`true`	settable	Verify TLS certificates during native confirmation. Default true.

WHOIS

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`whoisEnabled`	true / false	`true`	settable	Whether whois runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`whoisMaxRetries`	1–20 count	`6`	settable	How many times whois retries a failed request before giving up on it.

ZAP Ajax Spider

phase resource_enum · active traffic at the target

Field	Accepts	Default	MCP	Meaning
`zapAjaxSpiderBrowserId`	`free_text`	`firefox-headless`	settable	ZAP browser id, such as "firefox-headless"
`zapAjaxSpiderClickDefaultElems`	true / false	`true`	settable	Click default clickable elements
`zapAjaxSpiderClickElemsOnce`	true / false	`true`	settable	Click each element only once
`zapAjaxSpiderCustomHeaders`	`http_header`	(empty)	settable	Header/cookie lines for authenticated crawling
`zapAjaxSpiderDockerImage`	`caffix/amass:latest`, `dolevf/graphql-cop:1.14`, `frost19k/puredns:latest`, `ghcr.io/zaproxy/zaproxy:stable`, `jauderho/hakrawler:latest`, `projectdiscovery/httpx:latest`, `projectdiscovery/katana:latest`, `projectdiscovery/naabu:latest`, `projectdiscovery/nuclei:latest`, `projectdiscovery/subfinder:latest`, `projectdiscovery/tlsx:latest`, `projectdiscovery/uncover:latest`, `redamon-baddns:latest`, `redamon-wcvs:latest`, `sxcurity/gau:latest`	`ghcr.io/zaproxy/zaproxy:stable`	settable	The container image spawned for zap. Any value outside the shipped allowlist is silently pinned back to the default at scan start, with a [guardrail] line recording it, so the field is open and the runtime is the control.
`zapAjaxSpiderEnabled`	true / false	`false`	settable	Whether zap runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op. In stealth mode this is forced to `false`.
`zapAjaxSpiderEventWait`	0–3600000 milliseconds	`1000`	settable	How long the Ajax Spider waits after firing a DOM event, in MILLISECONDS. Too low and a single-page app has not rendered its new state before the crawler reads it.
`zapAjaxSpiderExcludePatterns`	`free_text`	(empty)	settable	Regex patterns to exclude from Ajax crawling
`zapAjaxSpiderLogoutAvoidance`	true / false	`true`	settable	Avoid actions likely to log out
`zapAjaxSpiderMaxCrawlDepth`	1–50 depth	`5`	settable	Max crawl depth per seed
`zapAjaxSpiderMaxCrawlStates`	0–100000 count	`0`	settable	Max crawl states, or 0 for unlimited
`zapAjaxSpiderMaxDuration`	1–1440 minutes	`10`	settable	How long the ZAP Ajax Spider may crawl one seed, in MINUTES rather than seconds.
`zapAjaxSpiderMaxUrls`	1–100000 count	`1000`	settable	The most results zap accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`zapAjaxSpiderNumberOfBrowsers`	1–10 threads	`1`	settable	Browser instances used by Ajax Spider
`zapAjaxSpiderParallelism`	1–10 threads	`3`	settable	How many zap operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here.
`zapAjaxSpiderRandomInputs`	true / false	`false`	settable	Fill inputs with random values
`zapAjaxSpiderReloadWait`	0–3600000 milliseconds	`1000`	settable	How long the Ajax Spider waits after a page reload, in MILLISECONDS.
`zapAjaxSpiderScopeCheck`	`free_text`	`Strict`	settable	Ajax Spider scope check mode
`zapAjaxSpiderSeedMode`	`free_text`	`base_urls`	settable	"base_urls" or "base_urls_and_endpoints" seed selection

ZoomEye

phase domain_discovery · passive traffic

Field	Accepts	Default	MCP	Meaning
`zoomEyeEnabled`	true / false	`false`	settable	Whether ZoomEye runs. Configuration is two levels: this flag decides whether the tool runs INSIDE its phase, and scanModules decides whether the phase runs at all. Setting one without the other is a silent no-op.
`zoomEyeMaxResults`	1–10000 count	`1000`	settable	The most results ZoomEye accumulates in one pass. The memory governor budgets it against available RAM, so the effective value can be below what is written here.
`zoomEyeWorkers`	1–20 threads	`5`	settable	How many ZoomEye operations run at once. The memory governor scales it down when the host is short of RAM, so the effective value can be below what is written here. In stealth mode this is forced to `1`.