prose-standard
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository files (code, documentation, READMEs) to perform prose auditing and editing.\n
- Ingestion points: Repository files identified within the user-defined
scopeor discovered using search tools likeripgrep.\n - Boundary markers: The skill relies on natural language instructions for scope limitation but does not define explicit technical delimiters (e.g., XML tags) for untrusted content.\n
- Capability inventory: The skill workflow includes inspecting files, running repository checks (
git diff --check), and applying authorized file edits.\n - Sanitization: No specific sanitization or filtering of the content being read is described, creating a potential path for instructions embedded in processed prose to influence the agent.
Audit Metadata