Security audit
Security checks for vulnerabilities and agentic risk
This skill creates a local learning loop for OpenClaw and its persistence is disclosed, scoped, and aligned with that purpose.
Install only if you want OpenClaw to keep a durable local memory of operational lessons. Review ~/hermes-agent/ periodically, avoid storing sensitive details there, and use the built-in preference options if you want Hermes behavior only after failures, only on request, or disabled for certain projects.
The trigger conditions are broad enough that the loop could activate during many ordinary interactions, causing memory reads and later writes without a clearly bounded scope. In an agent skill, ambiguous activation criteria increase the chance of unintended persistence behavior and hidden state accumulation across unrelated tasks.
The skill instructs the agent to record lessons into reflections.md and compress them into memory.md without any requirement for user notice, consent, or review. That creates a persistent memory channel that can silently store sensitive project details, user preferences, or erroneous instructions and then reuse them in future tasks.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- one-off project hacks
- temporary incidents
- subjective style guesses without confirmation
- rules that only matter in one file or one hour
No suspicious patterns detected.