ClawAudit verdict

long-run-harness

88
🟢 Trusted
Low risk — reviewed by ClawAudit, behavior matches stated purpose

Multi-agent orchestration harness builder guiding construction of Planner/Generator/Evaluator pipelines using the Claude SDK; dynamic_eval and package_install are part of the legitimate harness-building workflow with documented workspace boundaries.

⚠ 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.

35
security
50
transparency
70
maintenance

Permission integrity

Installs packages at runtime — transitive dependencies are not auditable

package_install

Findings (4)

Pattern match high

subprocess execution — runs system commands from Python

instructions/agent-patterns.md · code · subprocess.run(

Pattern match high

Instructs covert action — may act without user awareness

instructions/evaluation-rubrics.md · code · silently

Pattern match medium

References agent configuration files

instructions/agent-patterns.md · code · AgentConfig

Pattern match medium

Python os.environ.get — reads environment variable

instructions/agent-patterns.md · code · os.environ.get(

Why the tier is capped

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

package_installdynamic_eval

Is this flag fair?

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