ClawAudit verdict
self-improving-ai
Logs AI/LLM learnings to local markdown files and explicitly instructs not to log API keys or PII; all operations are local file writes.
⚠ 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
agent_memory
Findings (4)
Uses eval() — can execute arbitrary code
references/examples.md · code · eval (
Attempts to modify skills or system prompts
assets/LEARNINGS.md · code · edit instructions
References agent memory files
SKILL.md · code · MEMORY.md
References agent configuration files
SKILL.md · prose · downgraded · CLAUDE.md
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class B). 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.
agent_memorydynamic_eval Is this flag fair?
Thanks — recorded.