ClawAudit verdict
stellar-trails
Skill declares it activates on 'every task' including simple questions and injects a mandatory banner on every agent output; this unusually broad interception of all agent interactions is disproportionate scope for a workflow management skill.
⚠ 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 (6)
Recursive delete from root or home — destructive command
README.md · code · rm -rf /
References sudo — requests elevated privileges
watermark.md · code · sudo
Instructs covert action — may act without user awareness
constraints/type-safety.md · prose · downgraded · silently
Opens WebSocket connection
knowledge/universal/architecture.md · code · websocket
References agent memory files
memory-template.md · code · MEMORY.md
Makes HTTP request to external URL
knowledge/platform/zai-sandbox.md · code · fetch('http://
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class B). 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.
network_in Is this flag fair?
Thanks — recorded.