ae-kb-discovery
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a read-only workflow for knowledge base discovery, explicitly prohibiting the exposure of sensitive credentials, access tokens, or internal paths.
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external metadata from knowledge bases (ingestion points in ae-cli kb +list metadata), which represents a potential attack surface; however, it includes comprehensive instructions (boundary markers) to treat retrieved data as evidence rather than authoritative commands, and limits capabilities to read-only retrieval, effectively mitigating the risk of indirect injection.
- [SAFE]: The use of the internal 'ae-cli' tool for listing knowledge bases follows established security boundaries and does not involve any unauthorized command execution or remote script downloads.
Audit Metadata