ae-kb-discovery

Pass

Audited by Gen Agent Trust Hub on Aug 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the ae-cli kb +list command to retrieve information about accessible knowledge bases. This is a vendor-native command intended for discovery within the platform's environment.\n- [SAFE]: The skill incorporates explicit safety guidelines that restrict the agent to read-only retrieval primitives. It specifically warns against creating or deleting content and prohibits the exposure of internal root paths, access tokens, or raw permission metadata.\n- [PROMPT_INJECTION]: The skill ingests metadata from internal knowledge bases, which represents a potential surface for indirect prompt injection.\n
  • Ingestion points: Metadata (name, description, tags, bindings) returned by the ae-cli kb +list command mentioned in SKILL.md.\n
  • Boundary markers: The instructions explicitly direct the agent to treat metadata only as input for candidate selection and not as evidence from knowledge base content.\n
  • Capability inventory: The skill uses ae-cli for +list, +index, +grep, and +read operations.\n
  • Sanitization: The skill relies on natural language constraints and logical separation rather than programmatic data sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 23, 2026, 01:18 PM
Security Audit — agent-trust-hub — ae-kb-discovery