find-top-three
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external user-provided materials such as resumes, diaries, and financial data. While this creates a potential surface for indirect prompt injection, the instructions include robust mitigation strategies, such as requiring the model to strictly distinguish between 'confirmed facts' and 'user interpretations,' thereby preventing the agent from being misled by instructions embedded in external content.
- [PERSISTENCE]: The skill outlines a structured approach for creating and maintaining local markdown files to track user models and decision changes. This persistence is strictly scoped to documentation within the project folder and requires explicit user consent, differing from malicious persistence techniques that target system-level configuration or startup processes.
Audit Metadata