brandfy-mvp
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests user-controlled markdown from MVP.md and interpolates it into secondary documentation files like BRAND.md. This creates a surface for indirect prompt injection where adversarial instructions within the discovery document could be obeyed by subsequent agents analyzing the generated brand context.\n
- Ingestion points: The script
scripts/import-mvp.mjsreads content fromMVP.mdin the project root.\n - Boundary markers: Generated sections in the output files are delimited by
<!-- brandfy:mvpfy:start -->and<!-- brandfy:mvpfy:end -->tags.\n - Capability inventory: The script performs file system operations including reading, writing, and creating directories using
node:fs/promises.\n - Sanitization: The
plain()utility function removes markdown bolding, backticks, and HTML comments, but does not sanitize against malicious natural language instructions.
Audit Metadata