dbs-knowledge
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and summarizes local files, which creates a potential surface for indirect prompt injection if local data contains malicious instructions.
- Ingestion points: User-specified local directory files.
- Boundary markers: The skill uses a SOURCE_OF_TRUTH.md navigation file but does not define content-level delimiters.
- Capability inventory: File reading, listing, and writing/deletion (requires user confirmation).
- Sanitization: Instructions explicitly exclude sensitive paths such as .env, .git, and private keys.
- [SAFE]: No malicious patterns, unauthorized external network access, or obfuscation were identified. The skill follows best practices for local file manipulation and user confirmation.
Audit Metadata