ClawAudit verdict
multi-model-consensus
The skill facilitates multi-model decision-making, eliminating single-model biases and providing objective decision references. No malicious behavior or security risks are identified.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. ClawAudit flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Findings (4)
Writes to SKILL.md — self-modifying skill
docs/README.md · prose · downgraded · SKILL.md 新增 Read/Write
subprocess execution — runs system commands from Python
references/VERIFICATION_CASE.md · code · subprocess.run(
Accesses .ssh directory
SKILL.md · prose · downgraded · .ssh/
Popular HTTP library — network access
references/TROUBLESHOOTING.md · code · got
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class D). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
Is this flag fair?
Thanks — recorded.