ae-capability

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell-based interaction with the ae-cli tool to list, search, and run capabilities.
  • The agent is instructed to use commands like ae-cli capability run with JSON inputs, which is a standard operational pattern for this toolset.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection via the capability catalog it interacts with.
  • Ingestion points: Data enters the agent's context through ae-cli capability inspect and search commands in SKILL.md.
  • Boundary markers: The instructions do not define clear boundaries or specify that the agent should ignore instructions embedded in the metadata of external capabilities.
  • Capability inventory: The ae-cli capability run command allows for the execution of arbitrary actions defined in the catalog, which could include data writes or network requests.
  • Sanitization: There is no explicit requirement for the agent to sanitize or escape information retrieved from capability schemas before using that information to perform further actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 01:10 PM
Security Audit — agent-trust-hub — ae-capability