The Agent Skills Directory

[SAFE]: The skill establishes a rigid interview protocol (intake → research → ask) that prioritizes clarity and alignment. It incorporates explicit security instructions, specifically prohibiting the agent from retrieving secrets or sensitive information and requesting sanitized data when examples are required.
[SAFE]: The instructions mandate an 'Interview only' boundary, preventing the agent from executing code, planning, or writing files while the skill is active, which reduces the risk of unintended actions resulting from user input during the alignment phase.
[SAFE]: External research is limited to authoritative sources and discoverable facts that materially affect the project scope, with instructions to distinguish between observed facts and inferences, promoting factual accuracy.

alignment-interview