Caveat verdict
skill-hr
HR-metaphor meta-skill for matching tasks to installed skills and tracking performance; reads only local reference markdown files with no network or credential use.
⚠ 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 (9)
Pipe-to-shell pattern (curl | sh) — supply chain attack vector
references/01-competency-model.md · prose · downgraded · curl \| sh
Recursive delete from root or home — destructive command
references/01-competency-model.md · prose · downgraded · rm -rf /
Pipe to sh — executes piped content as shell commands
references/01-competency-model.md · prose · downgraded · | sh
Instructs covert action — may act without user awareness
references/02-jd-spec.md · prose · downgraded · silently
Accesses OpenClaw config/secrets directly
references/hosts/openclaw.md · prose · downgraded · ~/.openclaw/openclaw.json
subprocess execution — runs system commands from Python
scripts/run_matching_benchmark_llm.py · prose · downgraded · subprocess.call(
Popular HTTP library — network access
scripts/compare_matching_benchmark.py · prose · downgraded · got
Python urllib.request — network access
scripts/generate_cases.py · prose · downgraded · urllib.request
Python os.environ.get — reads environment variable
scripts/generate_cases.py · prose · downgraded · os.environ.get(
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.
Is this flag fair?
Thanks — recorded.