Caveat verdict
尸狗·警觉魄
shi-gou
A security sentinel module for detecting prompt injection, path traversal, and dangerous command patterns; purely defensive scanning with no exfiltration or suspicious execution.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Findings (6)
Prompt injection — tries to override agent instructions
README.md · code · ignore previous instructions
Raw model control tokens — prompt injection via token manipulation
README.md · prose · downgraded · <|im_start|>
Uses eval() — can execute arbitrary code
README.md · prose · downgraded · eval(
Possible prompt injection — attempts to redefine agent identity
README.md · prose · downgraded · you are now
Uses exec() — may execute shell commands
README.md · prose · downgraded · exec(
Sets world-executable permissions
scripts/security-check.js · prose · downgraded · chmod 777
Permissions & capabilities
No declared permissions — minimal attack surface.
Is this flag fair?
Thanks — recorded.