review-proposals
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes update suggestions generated by an external monitoring agent. If the source data or the monitoring agent itself is compromised, malicious instructions could be embedded in the suggestions.
- Ingestion points: The skill reads pending update suggestions from a monitoring agent's output file (SKILL.md, Step 2).
- Boundary markers: The skill requires displaying the full suggestion block (Step 3) and an exact diff (Step 4) to the user before action is taken.
- Capability inventory: The skill has the capability to write approved changes to the system file
~/.claude/plugins/config/claude-for-legal-zh/commercial-legal/CLAUDE.md. - Sanitization: The skill mandates that changes are only applied after a 'lawyer' (user) explicitly confirms the diff, providing a strong human-in-the-loop mitigation against malicious content.
Audit Metadata