to-docs
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from conversation history, code, and diffs to generate documentation and configuration changes. Ingestion points: SKILL.md indicates the agent reads conversation history, code, and diffs as the authoritative scope for decisions. Boundary markers: None identified; the instructions rely on the agent's reasoning to separate instructions from content. Capability inventory: The agent can edit documentation, update lint configurations, and modify tests. Sanitization: No explicit sanitization or escaping of input data is performed before it is used to suggest or apply documentation changes.
- [EXTERNAL_DOWNLOADS]: The skill references an external command for the user to install a dependency. Evidence: Phase 2 instructions suggest asking the user to run 'npx skills@latest add mattpocock/skills --skill writing-for-agents' if the writing-for-agents skill is missing.
Audit Metadata