improving-prompts
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and modify external prompt files (e.g., CLAUDE.md, AGENTS.md), which represents a surface for indirect prompt injection if those files contain adversarial instructions intended to influence the agent's behavior.
- Ingestion points: The skill specifically targets the optimization of user-provided prompt files and custom command instructions.
- Boundary markers: The skill mandates a "Required Process" in Step 1, requiring the agent to solicit concrete behavior issues and goals from the user before proceeding with any modifications. This acts as a critical human-in-the-loop validation boundary.
- Capability inventory: The agent utilizing this skill typically possesses file-system capabilities to read and write prompt configurations.
- Sanitization: The skill enforces strict adherence to documented guidelines rather than arbitrary execution of input content, although no automated technical sanitization is described.
Audit Metadata