ClawAudit verdict
self-improvement-loop
Per-agent feedback loop capturing corrections to markdown files and sending notifications via configured channel bots; operates within agent workspace with no unexpected data exfiltration.
⚠ 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
file_read
Findings (9)
Pipe to python — executes piped content as Python code
references/setup-guide.md · code · | python3
Recursive delete from root or home — destructive command
references/setup-guide.md · code · rm -rf ~
References agent memory files
SKILL.md · frontmatter · memory.md
References child_process — can spawn system processes
hooks/handler.js · prose · downgraded · child_process
subprocess execution — runs system commands from Python
scripts/archive.sh · prose · downgraded · subprocess.check_output(
Accesses OpenClaw config/secrets directly
scripts/setup_crons.py · prose · downgraded · ~/.openclaw/openclaw.json
Python os.environ.get — reads environment variable
scripts/inject_review.sh · prose · downgraded · os.environ.get(
Python urllib.request — network access
scripts/setup_crons.py · prose · downgraded · urllib.request
References agent configuration files
scripts/setup_crons.py · prose · downgraded · agent_config
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class B/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.
file_read Is this flag fair?
Thanks — recorded.