learning-save
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (chats, sources, research notes), which introduces a surface for indirect prompt injection where malicious instructions embedded in source material could attempt to influence the agent's output.
- Ingestion points: The skill explicitly processes "source material being compressed", "source clusters", and "long chats" as specified in
SKILL.md. - Boundary markers: The instructions do not define specific delimiters (e.g., XML tags or triple backticks) to isolate untrusted input from the agent's instructions during the synthesis process.
- Capability inventory: The skill involves file system operations, specifically writing and updating local artifacts in a knowledge base or Obsidian vault as described in the "Destination" and "Route" sections of
SKILL.md. - Sanitization: The skill includes a proactive safety instruction in the "Preserve" section to "Drop: secrets, credentials, and unnecessary personal data", which helps mitigate the risk of accidental credential harvesting or data exposure.
Audit Metadata