ClawAudit verdict
ima
tencent-ima-hermes
Tencent IMA knowledge base skill that clearly documents credentials are only sent to ima.qq.com and scoped COS endpoints, with explicit security disclosure in the frontmatter; behavior matches stated knowledge management purpose.
⚠ 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.
Permission integrity
credential_access
Findings (6)
Possible hardcoded credential
PUBLISH_GUIDE.md · code · APIKEY='*** /home/jerome/.hermes/skills/ima/bin/ima.sh list-kb | python3 -c
Pipe to python — executes piped content as Python code
PUBLISH_GUIDE.md · code · | python3
subprocess execution — runs system commands from Python
references/post-upload-validation.md · code · subprocess.run(
Pipe to bash — executes piped content as shell commands
references/post-upload-validation.md · prose · downgraded · | bash
subprocess with shell=True — command injection vector
SKILL.md · prose · downgraded · subprocess.run(["~/.hermes/skills/ima/bin/ima", "check-name", KB_ID, "A 股每日行情监控.
Accesses sensitive environment variables
scripts/build_kb_structure.py · prose · downgraded · process.env.IMA_OPENAPI_APIKEY
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class A/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.
credential_access Is this flag fair?
Thanks — recorded.