hunt-new-program — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited hunt-new-program (Agent Skill) and scored it 45/100 (orange). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 4 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 4 flagged
A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
*.sig, SIGNATURES) outside the documentation.A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
*.sig, SIGNATURES) outside the documentation.A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
*.sig, SIGNATURES) outside the documentation.A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
*.sig, SIGNATURES) outside the documentation.Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<context> Flags:
--auto - Use the provided brief as the primary source of truth and ask only for missing critical facts.</context>
<objective> Initialize a threat hunting program.
These hunt-native artifacts are the source of truth for the program.
Creates:
.planning/config.json.planning/MISSION.md.planning/HYPOTHESES.md.planning/SUCCESS_CRITERIA.md.planning/HUNTMAP.md.planning/STATE.md.planning/environment/ENVIRONMENT.md.planning/QUERIES/.planning/RECEIPTS/Bootstrap should only scaffold the program. Do not seed sample queries, sample receipts, or completed phases. Unknown environment facts, tools, retention windows, and owners must remain TBD unless the operator confirms them. Confirmed bootstrap facts such as the program name, mode, opened date, and initial phase/status must be filled immediately.
After this command: Run /hunt-map-environment to capture confirmed facts, or edit .planning/environment/ENVIRONMENT.md manually and continue later. </objective>
<execution_context> @.github/thrunt-god/workflows/hunt-bootstrap.md @.github/thrunt-god/templates/config.json @.github/thrunt-god/templates/mission.md @.github/thrunt-god/templates/hypotheses.md @.github/thrunt-god/templates/success-criteria.md @.github/thrunt-god/templates/hunt-program-huntmap.md @.github/thrunt-god/templates/hunt-state.md @.github/thrunt-god/templates/environment-map.md </execution_context>
<process> Execute the bootstrap workflow from @.github/thrunt-god/workflows/hunt-bootstrap.md in program mode. Drive the conversation through .planning/environment/ENVIRONMENT.md and the operator toolchain before defining later hunt phases. Create .planning/QUERIES/ and .planning/RECEIPTS/ as empty directories only. Do not load query-log or receipt templates during bootstrap; those belong to /hunt-run after real execution begins. Default behavior is scaffold-first: write confirmed facts only and leave unknown values as TBD instead of inventing sample content. Create .planning/config.json during bootstrap if it does not already exist so runtime, settings, and connector commands are immediately usable. Never hand-write .planning/config.json; use thrunt-tools config-new-program and thrunt-tools config-set so the file stays valid THRUNT config. Use built-in connector ids exactly as the runtime registers them, for example splunk and elastic; do not substitute elasticsearch. When writing connector profiles, use base_url for the runtime URL field; do not invent or substitute endpoint. Only configure connector profiles when auth type and secret ref names are confirmed. Never invent placeholder env vars or placeholder secrets for blocked connectors. When writing secret_refs, each confirmed secret must use the THRUNT object shape { "type": "env", "value": "ENV_VAR_NAME" } rather than a raw string. Keep connector narrative, status notes, and access commentary in ENVIRONMENT.md, not in ad hoc config keys. Do not leave bootstrap-known fields as TBD after writing the files. Write the hunt artifacts directly. Preserve any existing user-authored content unless the user explicitly wants a reset. </process>
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.