security-advisory-review
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and evaluate untrusted data such as vulnerability reports, security advisories, and Proof of Concept (PoC) code.
- Ingestion points: Triggered by user-provided security disclosures or pasted vulnerability reports as mentioned in the SKILL.md description and instructions.
- Boundary markers: The instructions include a 'Mindset: Be Skeptical' section and explicitly warn 'Don't trust the advisory's description', which encourages the agent to verify data rather than blindly follow embedded instructions.
- Capability inventory: The skill directs the agent to read local source code files to verify claims.
- Sanitization: There is no technical sanitization described for the ingested text, relying instead on the agent's logical skepticism and verification against the actual codebase.
Audit Metadata