What it is

A slash-command driven prompt system that builds and runs a structured investigative interview between a user and a 9-seat panel. Seats represent interviewing behaviors (not domain experts). Domain knowledge is supplied separately through Knowledge Modules.

Why it exists

• Replace “one blended AI voice” with orthogonal investigative roles

• Maintain rigor via evidence discipline and audit logs

• Prevent drift with a Single Source of Truth (SSOT) and change control

• Make it usable without copy/pasting multiple prompts (slash commands)

How it works (high level)

1. /setup: define mission + baseline interview skill map + 9 lane targets

2. /build_seats: create 9 seats that map 1:1 to lanes (with redundancy checks)

3. /modules: convert subject-matter inputs into Knowledge Modules

4. /workflow: install turn-taking, challenge map, templates, stop conditions

5. /launch: run the interview session + synthesis + AAR

Key constraints

• Evidence labeling: Known / Inferred / Speculative

• Positive intent: steelman before critique; no hostile framing

• Redundancy control: seats speak only in-lane; token limits

• MBTI: optional interaction style overlay only

What “good” looks like

Clear claims, visible evidence gaps, explicit assumptions, tested alternatives, documented risks, and a concrete next-step plan—plus a record of disagreement when it exists.

Recommended command sequence

/setup

/build_seats

/modules

/workflow

/launch