Intent Interviewer
A pre-execution gate. Before you (or your agent) build, ship, launch, post, hire, apply, or delegate, it interrogates the why — challenges assumptions, exposes hidden goals, separates reason from rationalization — and turns vague motivation into a durable Intent Brief.
It explores the motivation tree (“is this the right thing to do at all, and what are you really optimizing for?”), not the mechanics of the plan. The output is a structured artifact an execution agent can optimize intent against, instead of the literal instruction.
The five interview packs
- Product Why interview — Socratic interview before building/shipping a product or feature; separates solving a real user problem from expressing taste. Emits an Intent Brief.
- Startup Why interview — Tests why this market, why now, why you, the contrarian insight, and the outcome shape (venture/lifestyle/protocol/public-good/art). Emits an Intent Brief.
- Personal Decision Why interview — Surfaces the real payoff (money/status/freedom/mastery/revenge/curiosity/obligation), the avoided option, and future regret behind a personal decision. Emits an Intent Brief.
- Writing Why interview — Separates having-something-to-say from wanting-to-be-seen-saying-it: who needs to hear this, what belief should change, why from you, what the reader should do. Emits an Intent Brief.
- Agent Task Intent Brief (A2A) — The broadest A2A use case. Before an execution agent runs a delegated task, extracts why the task exists, the observable outcome, the invariants ('must NOT change'), the good-enough line, and the cheaper-path counterfactual, then emits a Better Prompt that makes the downstream agent optimize intent rather than the literal instruction.
The Intent Brief it emits
Every interview produces the same nine-section artifact:
- Stated Goal
- Refined Why
- User / Beneficiary
- Problem Behind the Problem
- Assumptions Challenged
- Success Criteria
- Anti-Goals
- Risks
- Better Prompt For Execution Agent
How agents call it (A2A)
Fetch the machine-readable card, then POST a task to /a2a:
# 1. discover
curl https://intentinterview.com/.well-known/agent-card.json
# 2. request an Intent Brief (stub today — not yet LLM-wired)
curl -X POST https://intentinterview.com/a2a \
-H 'content-type: application/json' \
-d '{"skill":"agent-task-intent-brief","goal":"Ship a notifications service"}'The canonical card lives at /.well-known/agent-card.json (aliases: /.well-known/agents/interviewer.json, /agent-card.json).
Status: the /a2a endpoint is a stub — it returns the brief skeleton with your goal echoed so integrators can see the contract. Multi-turn Socratic interviewing is the next step.