do-your-job
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides instructions that prioritize user-provided 'diagnoses' and 'directives' over the agent's internal safety and logic systems.
- Evidence: The skill commands the agent to "do the stated thing verbatim" and states that "the user's diagnosis outranks your preference."
- Ingestion points: The agent context is updated via user input identified as "corrections", "diagnoses", or "directives" (e.g., "it just needs X").
- Boundary markers: Absent. There are no instructions to validate the safety or feasibility of these directives before verbatim implementation.
- Capability inventory: The skill implies access to file editing, document updates, and general tool execution through its association with
agent:obey. - Sanitization: Absent. The skill explicitly forbids "embellishment" or "theorizing," which effectively bypasses the agent's typical output sanitization and safety reasoning steps.
Audit Metadata