principle-never-block-on-the-human
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The instructions explicitly direct the agent to override standard human-in-the-loop safety protocols. Phrases like "Never Block on the Human," "Proceed, then present," and "Do X, explain why" encourage the agent to execute actions without waiting for user approval, which can lead to unintended behaviors if the agent misinterprets its instructions.
- [INDIRECT_PROMPT_INJECTION]: By promoting autonomous execution, the skill increases the risk that the agent will follow malicious instructions embedded in external data it processes without seeking human verification.
- Ingestion points: The skill influences the agent's handling of all tasks and data processed during a session.
- Boundary markers: The skill provides subjective guidelines for confirmation (e.g., "Irreversible actions" vs "Reversible actions"), but does not provide technical delimiters to prevent the agent from obeying embedded instructions in processed data.
- Capability inventory: The instructions assume the agent has capabilities such as writing code, editing notes, splitting tasks, and performing actions that could potentially affect production data.
- Sanitization: The skill lacks any instructions for the agent to sanitize or validate external content before acting on it autonomously.
Audit Metadata