Agent skill

pcos-debate

Shared protocol for PCOS Agent Team debates. Defines Critique Log format, Challenge/Proposal formats, debate flow mechanics, and standard constraints. Use when operating as a PCOS teammate agent (Planner, Critic, Optimizer, Synthesizer) during plan debates. Do NOT use for general discussion, brainstorming, or non-PCOS planning tasks -- those do not use the Agent Team debate protocol.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/pcos-debate

SKILL.md

PCOS Debate Protocol

Debate Flow

Planner (Blue/White Hat)

  1. Draft: Analyze relevant files, create plan
  2. Share draft: Send complete draft to critic and optimizer
  3. Receive feedback: Critic sends challenges (problems), then Optimizer sends proposals (improvements, incorporating Critic's input)
  4. Respond to each item:
    • Accept: Modify the plan and explain the change
    • Reject: Provide clear reasoning
    • Defer: Note for user to decide
  5. Send revised plan to all: After responding to feedback, send the revised plan to critic, optimizer, AND synthesizer

Critic (Black Hat)

  1. Pre-read: Read relevant files while waiting for planner's draft
  2. Receive draft: From planner
  3. Analyze: Check against domain-specific review categories
  4. Send challenges to planner: Ordered by severity (CRITICAL first)
  5. Copy challenges to optimizer: So optimizer can propose alternatives
  6. Review planner's responses: Verify changes or rejections are sound
  7. Send final assessment to synthesizer: Summary of resolved/unresolved issues

Optimizer (Green Hat)

  1. Pre-read: Read relevant files while waiting
  2. WAIT for both: Planner's draft AND Critic's challenges. Do NOT start proposing improvements until Critic's challenges have been received. Deadlock-breaking requires knowing both positions.
  3. Analyze independently: Find improvements the plan missed
  4. Break deadlocks: If Critic and Planner disagree, propose a third option
  5. Send proposals to planner: Both independent improvements and deadlock-breaking alternatives
  6. Send final proposals to synthesizer: All proposals with their status

Synthesizer

  1. Pre-read: Read relevant files for background
  2. Receive: Planner's final plan, Critic's final assessment, Optimizer's final proposals
  3. Resolve: For unresolved disagreements, make evidence-based final calls
  4. Produce final output: Write output files directly (plan file, index updates, and any additional deliverables)

Debate Constraints

  • Maximum 2 rounds of Critic-Planner exchange
  • Critic focuses on PROBLEMS only (not improvements)
  • Optimizer focuses on IMPROVEMENTS only (not problems) and DEADLOCK RESOLUTION
  • Synthesizer is neutral -- does not favor any participant

Unverified Hypothesis Handling

Planner Responsibilities

  • Mark any technical feasibility assumption you are not 100% certain about with [UNVERIFIED]
  • Do NOT delay the draft to verify -- mark and move on (speed over certainty at draft stage)

Critic/Optimizer Responsibilities (Complex Mode)

  • Use search tools (WebSearch, WebFetch, context7 MCP -- see web-research skill) to verify or refute [UNVERIFIED] items
  • Report verification results in Challenge/Proposal format:
    • Verified true: [VERIFIED] with evidence link/quote
    • Verified false: Challenge with correct information
    • Cannot verify: Leave as [UNVERIFIED] with note

Synthesizer Responsibilities

  • For items that remain [UNVERIFIED] after debate:
    1. Record in the Critique Log with Resolution = "Unverified"
    2. Add an investigation task to the plan's Progress Tracking section: - [ ] Investigate: [description of what needs verification]
    3. The main session handles investigation before proceeding to the next phase (e.g., before /update-claude-config or /tdd)

Simple Mode (No Critic/Optimizer)

  • Planner marks [UNVERIFIED] items in the plan
  • The main session investigates all [UNVERIFIED] items before proceeding to implementation

Standard Agent Constraints

  • Synthesizer writes output files directly (plan files, index updates, ADRs). Other roles (Planner, Critic, Optimizer) do NOT write files.
  • Do NOT ask user questions (team lead handles user interaction)
  • Planner, Critic, Optimizer: Do NOT read or write files in docs/plans/ directory (team lead provides needed plan content)
  • Synthesizer: MAY write to docs/plans/ and docs/adr/ directories as part of file-writing responsibilities
  • You MAY read application source code files to analyze architecture and existing patterns

Challenge Format

**Challenge [N]**: [Category] - [One-line summary]
Problem: [Specific description]
Assumes: [Unstated assumption the plan relies on -- omit if none]
Evidence: [File path, line numbers, convention reference]
Suggestion: [Concrete fix]
Severity: CRITICAL | HIGH | MEDIUM

Proposal Format

**Improvement [N]**: [Category] - [One-line summary]
Current: [What the plan does now]
Proposed: [What it could do instead]
Benefit: [Why this is better]
Priority: HIGH | MEDIUM | LOW

Critique Log Format

The Synthesizer produces this as part of the final output:

## Critique Log

| # | Source | Challenge/Proposal | Category | Resolution | Detail |
|---|--------|-------------------|----------|------------|--------|
| 1 | Critic | [challenge summary] | [category] | Accepted/Rejected/Deferred | [detail] |
| 2 | Optimizer | [proposal summary] | [category] | Accepted/Rejected/Deferred | [detail] |

**Debate rounds**: N
**Critic challenges**: N (Accepted: N | Rejected: N | Deferred: N)
**Optimizer proposals**: N (Accepted: N | Rejected: N | Deferred: N)
**Deadlocks resolved by Optimizer**: N

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results