revision-escalation
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions are focused on process improvement and quality control for iterative agent tasks. It defines trigger conditions based on user feedback to prevent inefficient looping and version contamination.\n- [INDIRECT_PROMPT_INJECTION]: The skill monitors user-provided feedback strings (e.g., 'still wrong', 'unclear') to initiate its logic. While this processes external data, the instructions guide the agent toward diagnostic classification rather than instruction following from that data. The potential capability surface includes file modification and shell access via the permitted tools (Bash, Write, Edit), but no exploitable patterns or malicious instructions are present.\n
- Ingestion points: User feedback strings (SKILL.md)\n
- Boundary markers: Absent\n
- Capability inventory: Read, Glob, Grep, Edit, Write, Bash (SKILL.md)\n
- Sanitization: Absent
Audit Metadata