- Authorized scanning only: Run the scanner only against assets you own or have explicit permission to test.
- Scope controls: Configure
scanner.scope.include_domains/allowlist; SSRF guard blocks hosts outside scope and private IPs unlessallow_private=true. - Data handling: Findings/evidence may contain sensitive data. Store reports securely and avoid sharing outside trusted channels.
- Browser automation: Playwright-driven verification may execute target scripts. Use isolated environments and keep browsers patched.
- Evidence-first interpretation:
verifiedmeans the current implementation captured stronger, reproducible proof.detectedmeans a concrete signal was observed, but exploit confirmation is incomplete.suspectedmeans the behavior was interesting enough to report, but not strong enough to call confirmed.informationalmeans the scanner observed posture or inventory data that may matter operationally but is not itself an exploit confirmation.- Anything not evidenced should be treated as untrusted and manually reviewed before action.
- Anything not reproducible must not be treated as confirmed.
- Auth harness:
- Access-control findings must include cross-actor evidence before they can remain
verified. - Actors are not considered ready unless login, verify, or refresh produced explicit proof.
- Browser-driven login does not count as successful unless it yields a usable session state and HTTP or verification proof.
- Placeholder credentials in
config/default_config.yamluse environment variable references only; do not commit real secrets. - If login or refresh fails, the affected actor is reported as degraded/login_failed/refresh_failed and related verification scenarios remain partial.
- Browser actor artifacts are optional supporting evidence, not the sole basis for confirmation.
- Deterministic RBAC verification requires an explicit policy matrix plus actor readiness; otherwise the engine stays heuristic or partial by design.
- Replay results are evidence aids, not independent proof, unless they reproduce the actor-scoped scenario truthfully.
- Workflow verification requires step-level evidence and checkpoint agreement; a workflow marked
partial,failed,blocked_auth, orindeterminatemust not be treated as a completed verification. - Workflow replay is separate from request replay and only counts as reproduced when the replayed workflow matches the original checkpoint trail.
- Access-control findings must include cross-actor evidence before they can remain
- Local/mock labeling: Reports classify localhost/private targets as local test environments so demo or smoke results are not mistaken for production findings.
- Disclosure: Report vulnerabilities responsibly to the affected parties following their disclosure policy.
If you discover a security issue in this project, open a private issue or contact the maintainers with a minimal, reproducible description.