derive-section-voice-overlay
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from published posts which could contain adversarial content designed to influence the AI's analysis or output.
- Ingestion points: Reads content from
corpus/published/{section}/as part of its pattern extraction workflow (SKILL.md). - Boundary markers: The workflow does not specify the use of delimiters or instructions to ignore embedded commands within the ingested posts.
- Capability inventory: The skill generates and writes markdown files to the
shared-context/voices/directory based on the analyzed data (SKILL.md). - Sanitization: There is no mention of sanitizing or filtering the input data before it is processed and written to the draft overlay file.
Audit Metadata