consult-outside-expert
Pass
Audited by Gen Agent Trust Hub on Mar 8, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted 'Artifacts' and 'External Review' feedback as part of its core loop, creating an indirect injection surface.
- Ingestion points: Artifacts (files) and external expert responses entering the agent context as described in
references/guide.md. - Boundary markers: The guide uses markdown headers and templates to structure inputs, but lacks explicit instructions to the agent to ignore or escape instructions contained within the reviewed content.
- Capability inventory: The agent has the capability to read/write local files and execute shell commands.
- Sanitization: There is no evidence of sanitization, filtering, or validation of the content of the artifacts or the feedback before processing.
- [COMMAND_EXECUTION]: The
references/guide.mdfile provides a suite of shell commands (usinggrep,test,wc) for the agent to execute in order to verify the structure and existence of session logs.
Audit Metadata