inversion
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates as a structured prompt for cognitive analysis and does not perform network operations, access sensitive files, or execute system commands.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied plans by interpolating them into its prompt template. (1) Ingestion points: User input for plans in SKILL.md. (2) Boundary markers: Absent. (3) Capability inventory: Text generation only (mentions calling other analytical skills kaz-leader and first-principles). (4) Sanitization: Absent. The lack of boundary markers is a minor best-practice observation with no significant risk given the lack of dangerous tool access.
Audit Metadata