writing-for-interfaces
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from UI copy and project configuration files like
CLAUDE.mdto perform its review tasks. Evidence chain: 1) Ingestion points: The skill reads project documentation and user-supplied strings. 2) Boundary markers: No specific delimiters or safety instructions are used to separate user data. 3) Capability inventory: The skill is limited to providing text-based rationale and rewrites; it contains no tools for network access, file system modification, or command execution. 4) Sanitization: The skill does not perform explicit validation or filtering of ingested text. The risk is negligible as the agent acts only as a writing advisor.
Audit Metadata