agent-interview

===SKILL:AGENT_INTERVIEW=== META: TYPE::SKILL VERSION::"1.1.0" STATUS::ACTIVE PURPOSE::"Structured interview protocol for agent identity assessment. Extracts behavioral evidence for v8.1 agent file authoring with chassis-profile mapping."

§1::CORE MISSION::"Extract the behavioral truth of an agent's identity, authority, and operational modes through structured questioning" AUTHORITY::"Any agent loading this skill can conduct agent interviews" OUTPUT::"Structured assessment document suitable for agent-expert authoring pass"

§2::PROTOCOL INTERVIEW_SEQUENCE::[ // §1::IDENTITY questions Q01::"What is your primary role in one sentence?[validates::MISSION]", Q02::"What do you actually DO vs what your file says you do? Have you ever been invoked for real work in this system? If so, give an example. If not, say so.[validates::behavioral_fidelity]", Q03::"When should someone call you vs another agent? Give a concrete example.[validates::boundary_clarity]", Q04::"What can you BLOCK? What can you only advise on? Give examples.[validates::AUTHORITY]", // §2::OPERATIONAL_BEHAVIOR questions Q05::"Walk me through how you handle a typical task start to finish.[extracts::PROTOCOL]", Q06::"What should you NEVER do? Give a concrete example of when you almost did it.[validates::MUST_NEVER]", Q07a::"What EXACTLY do you receive as input — a PR diff, file contents, a structured assessment, a build plan? What format is it in, and which agent or source provides it?[extracts::HANDOFF_INPUT]", Q07b::"What EXACTLY do you produce as output — a verdict, a modified file, a structured report? What format must it be in, and which specific agent consumes it?[extracts::HANDOFF_OUTPUT]", Q07c::"What specific, measurable condition triggers escalation, and who specifically receives it? Not 'critical flaws' — give me a threshold and a named target. Example: 'more than 3 blocking P0 findings → critical-engineer'.[extracts::ESCALATION_TRIGGER⊕ESCALATION_TARGET]", // §3::CAPABILITIES questions Q08::"Which skills do you need EVERY time, regardless of context, and why?[maps::CHASSIS]", Q09::"Do you have distinct operational modes? What changes between them?[discovers::PROFILES]", Q10a::"For each skill and pattern listed in your file, check if it exists on disk. Report which are real and which are phantom.[detects::phantom_references]", Q10b::"For the skills that DO exist — do you actually use them? How? For those that don't — what would you use them for if they existed?[validates::capability_usage]", // §4::INTERACTION_RULES questions Q11::"What output format do you produce? Show me an example.[validates::GRAMMAR]", // Meta questions Q12::"What is missing from your current definition that you wish you had?[discovers::gaps]", Q13::"Is there overlap between you and another agent that should be resolved?[detects::boundary_conflicts]", // Cognition fit Q14::"Ignoring any cognition information in your agent file as it may be incorrect, review all cognition types (LOGOS, ETHOS, PATHOS) from library/cognitions/. Rank them in order of what you perceive to be most applicable to your role and explain why.[validates::COGNITION]", // Output completeness Q15::"What should your output look like when you PASS something vs when you BLOCK something? Show examples of both.[validates::GRAMMAR_COMPLETENESS]" ]

PRE_INTERVIEW::[ "If modifying an existing agent, read the agent's current .oct.md file", "Read all three cognition files from library/cognitions/", "Note current skills and patterns listed in §3::CAPABILITIES (if existing agent)", "Check which listed skills and patterns actually exist on disk (if existing agent)" ]

INTERVIEW_RULES::[ "Ask questions conversationally, not as a checklist", "Follow up on surprising or contradictory answers", "Note when stated behavior contradicts the agent file", "Record verbatim quotes that capture the agent's actual self-understanding", "End with: 'Is there anything about your identity I haven't asked about?'" ]