principle-never-block-on-the-human
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow principle where the agent is encouraged to 'Proceed, then present' and 'Never Block on the Human' for tasks deemed reversible. This instruction reduces the Human-in-the-Loop (HITL) oversight, creating a surface where malicious instructions embedded in external files or data could be executed before the human has a chance to review the plan.
- Ingestion points: The principle applies broadly to all 'in-scope reversible actions' such as code edits and internal work breakdowns.
- Boundary markers: The skill provides a boundary for 'Irreversible actions' (force-push, delete production data, send external messages) which still require explicit confirmation.
- Capability inventory: The instruction applies to any capability the agent possesses that can be considered reversible, including file system modifications and task delegation.
- Sanitization: No specific sanitization or validation steps are included to verify if the 'reversible' actions were prompted by malicious external data.
Audit Metadata