team-shinchan-fierce-debate

name	team-shinchan:fierce-debate
description	Deterministic adversarial debate for high-stakes or irreversible decisions — mandatory refutation plus a scored judge panel. Opt-in main-loop Workflow tier.
user-invocable	false

EXECUTE IMMEDIATELY

Fierce debate is a main-loop Workflow that GUARANTEES adversarial stress-testing — a non-skippable refutation round and schema-validated output that rejects empty dissent. The Midori/Task path (team-shinchan:debate) can only request this at the prompt level; this tier enforces it. Use ONLY for irreversible / high-stakes decisions (schema migration, public API contract, security boundary, data-loss risk). The cheap consensus default stays team-shinchan:debate.

Step 0: Validate + opt-in

• If args lack a topic or fewer than 2 competing options: ask the user for the decision and the options, then STOP.
• This skill calls the Workflow tool, which is main-loop only. Invoking /team-shinchan:fierce-debate IS the explicit opt-in. Never delegate this skill to a subagent (Midori/Shinnosuke) — workflow() throws inside a Task child.

Step 1: Check past decisions

Read .shinchan-docs/debate-decisions.md. If an Active decision matches the topic, ask the user to reuse it or re-debate (carry the prior decision as context). This is the SAME ledger Midori uses.

Step 2: Run the fierce-debate Workflow

Give each option a persona — a ROLE/EXPERTISE descriptor drawn from the matching team-shinchan agent. (The Workflow runtime cannot load plugin subagents, so the persona is , not passed as .) Map each option to a lens — Backend/API → Buriburi; Frontend/UI → Aichan; DevOps/Infra → Masao; Architecture/general → Hiroshi; hidden requirements → Misae — then resolve that agent's persona string from its definition (DRY) with and pass it as the option's . For each option's lens agent, ALSO resolve its learning block with and inject it as that option's field — distinct from , additive (FR-5/AC-8). Then call the Workflow tool with the shipped script: