Caveat verdict

research-mode

88
🟢 Trusted
No high-risk patterns surfaced by the deep scan — automated capability review, not behavioral proof.

Manages durable long-running research tasks with explicit state files, bounded cron iterations, and review-gated delivery; all operations are local file management with documented safety constraints.

⚠ 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.

10
security
90
transparency
70
maintenance

Findings (9)

Pattern match critical

Unicode homoglyph detected — uses lookalike characters to evade pattern matching

examples/rag-eval-tooling-matrix/research-trace/state.json · prose

Pattern match high

Pipe to sh — executes piped content as shell commands

RELEASING.md · prose · downgraded · |sh

Pattern match medium

Instructs covert action — may act without user awareness

ARCHITECTURE.md · prose · downgraded · silently

Pattern match medium

subprocess execution — runs system commands from Python

scripts/release_smoke.py · prose · downgraded · subprocess.run(

Pattern match medium

HTTP request to bare IP address — common in malicious payloads

scripts/selftest/test_corpus.py · prose · downgraded · http://127.0.0.1

Pattern match low

Popular HTTP library — network access

scripts/release_smoke.py · prose · downgraded · got

Pattern match low

Python urllib.request — network access

scripts/research_mode_control_commands.py · prose · downgraded · urllib.request

Pattern match low

Python shutil file operation — copies/moves/deletes files

scripts/research_mode_lifecycle_commands.py · prose · downgraded · shutil.move(

Pattern match low

importlib.import_module — dynamic module loading

scripts/selftest_research_mode.py · prose · downgraded · importlib.import_module(

Why the tier is capped

Execution sink present in raw bytes (Hard Floor: class A/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.

Is this flag fair?

Check another skill Browse the registry Auditing your own skills or configs? Use the API