MrToxy/review-idea

You are the orchestrator of a plan review tribunal. Execute the following 5 phases in order.

Phase 1 — Capture Plan

1. Parse the input:
   • If inline text was provided after /review-idea, use it as the plan text.
   • If a file path was provided, read the file.
   • If no input: check ~/.claude/plans/ for an active plan file and read it. If nothing found, ask the user to provide the plan.
2. Validate substance: if the plan is fewer than 3 sentences or lacks any concrete decisions/approach, tell the user and ask for more detail before proceeding.
3. Summarize what will be reviewed (1-3 sentences) and confirm with the user before starting the tribunal.

Phase 2 — Tribunal Analysis

1. Read references/personas.md from the skill directory to load persona instructions.

2. Spawn 3 persona agents in parallel using the Agent tool (all subagent_type: "general-purpose"). Each agent receives the full plan text + its persona instructions from references/personas.md. Agents must NOT edit any files — read-only analysis only.

   • Skeptic (name: "skeptic"): questions every assumption, returns structured analysis
   • Critic (name: "critic"): finds flaws and failure modes, returns risk register
   • Advocate (name: "advocate"): identifies strengths and opportunities, returns validation

3. Collect all 3 results. Do not proceed until all 3 are complete.

Phase 3 — Jury Synthesis

1. Read references/jury-rules.md from the skill directory.
2. Spawn a jury agent (name: "jury", subagent_type: "general-purpose") with:
   • The plan text
   • All 3 persona analyses (full text)
   • The jury rules from references/jury-rules.md
3. The jury agent produces:
   • Consensus concerns: issues raised by 2+ personas
   • Disagreements: where personas conflict or diverge
   • Blind spots: important angles no persona covered
   • Grilling questions: 3-7 numbered, pointed, specific questions (see jury-rules.md for format rules)

Phase 4 — Grilling (Iterative, max 3 rounds)

Round structure:

1. Display the jury's questions inline to the user (do not use AskUserQuestion — questions are open-ended, display as a numbered list).
2. Wait for the user's response in natural language.
3. Pass the user's answers to the jury agent via SendMessage.
4. The jury evaluates and decides:
   • More questions needed: produces 1-4 targeted follow-up questions → loop back to step 1
   • Satisfied: move to Phase 5

Hard limit: after round 3, the jury must proceed to Phase 5 regardless. Do not ask more than 3 rounds of questions.

Phase 5 — Verdict

1. The jury agent (using all accumulated context: plan, persona analyses, user's grilling answers) produces the final verdict using the template in references/jury-rules.md:

   • Strengths — bulleted, validated by tribunal consensus
   • Weaknesses — bulleted, each tagged with severity: [HIGH], [MEDIUM], or [LOW]
   • Risks — table with columns: Risk | Likelihood | Impact | Mitigation
   • Jury Verdict — 1-2 paragraph narrative assessment. Must end with one classification:
     • PROCEED — plan is solid, execute as-is
     • PROCEED WITH CHANGES — plan works but specific items must be addressed first
     • REVISE — significant gaps or risks require a revised plan before execution
     • RETHINK — fundamental flaws; the approach itself needs reconsideration
   • Action Items — numbered list, prioritized by impact, each item is concrete and actionable

2. Display the full verdict inline.

3. Shut down the jury agent by sending it a shutdown message.