audit-harness-safety
Skill7 starsChanged 49 days agoNo licenceView on GitHub
Audit an agent harness's setup for blast-radius containment and recovery — OS-level sandbox + least-privilege scope, hard deny floors, reversibility, a kill path, and consumption bounds — bounding the *damage* if a control fails. Invoke when hardening a harness for destructive-action safety — before granting write/Bash tools, running an agent headless or in CI, or when it touches the filesystem, real credentials, shared branches, prod, or paid APIs; also reviewing whether damage/spend is bounded if a run goes wrong. Skip when the question is exfiltration / whether a web-access grant is safe (use audit-lethal-trifecta), trusting an LLM-emitted string or agent-run install at the sink — even asked as a sandbox question (use audit-supply-chain-sinks), workflow-YAML injection/token/pin risks (use audit-github-actions-security), a stated rule's enforcement medium — prose vs hook vs deny rule (use audit-prompt-hook-placement), or run visibility — traces, is the runaway auto-halted (use audit-observability-setup).
- Reads credentials
In this repository24 other files it gives its agents.
- AGENTS.mdAGENTS.md
- architecture-committeeSkill
- audit-geo-contentSkill
- audit-github-actions-securitySkill
- audit-instruction-fileSkill
- audit-lethal-trifectaSkill
- audit-mcp-serverSkill
- audit-memory-retrieval-integritySkill
- audit-multi-agent-orchestrationSkill
- audit-observability-setupSkill
- audit-prompt-cache-hygieneSkill
- audit-prompt-hook-placementSkill
- audit-secret-exposureSkill
- audit-self-improvement-protocolSkill
- audit-skill-qualitySkill
- audit-supply-chain-sinksSkill
- audit-tool-definitionSkill
- audit-verification-gatesSkill
- compress-promptSkill
- conflicting-instruction-detectorSkill
- review-codeSkill
- write-skillSkill
- write-testsSkill
- write-tool-descriptionSkill
The file
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.