lesson-to-skill
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (runtime logs, code snippets, and user reports) to update the agent's permanent instruction files. This creates an attack surface for indirect prompt injection where malicious content in logs could influence future agent behavior.
- Ingestion points: The skill ingests data from sources labeled
[code],[runtime],[ticket],[user], and[skill](SKILL.md). - Boundary markers: The skill includes a 'Verification Gate (Hard Stop)' that requires reproducible proof and explicit approval before modifying any files, which significantly reduces the risk of accidental or malicious rule poisoning.
- Capability inventory: The skill utilizes file-writing capabilities to 'Patch the Owner' (Step 3) and 'Create' new skill files (Step 4) within directories like
.github/skills/,.cursor/rules/, and root instructions. - Sanitization: The instructions mandate that lessons be formulated as single sentences and specifically forbid the inclusion of session logs, incident narratives, or speculative rules, serving as a structural sanitization method.
Audit Metadata