plan-debate
Skill1 starsChanged 7 months agoMITView on GitHub
Multi-agent debate planning pipeline for high-quality implementation plans. Uses agent separation and independent review to catch issues before code is written. For simple tasks (1-2 steps, no dependencies), plan inline without this skill. Every plan must include a Verification Plan with applicable tiers: Research-only plans (no code changes) skip verification tiers entirely. When a plan touches UI, four phases are mandatory:
1 other project uses a version of this.
Adopted by
Most recently taken up by
- zirui-song/claude-skillstheir own version, first seen Nov 1, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- zirui-song/claude-skills★ 17same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository14 other files it gives its agents.
- brew-maintainSkill
- cc-updatesSkill
- commit-helperSkill
- dev-consoleSkill
- explain-codeSkill
- init-modularSkill
- learnSkill
- port-managerSkill
- product-contextSkill
- projectSkill
- refactorSkill
- techdebtSkill
- ui-specSkill
- web-researcherSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 51
Every plan must include a Verification Plan with applicable tiers:
- Distinctive: line 65
When a plan touches UI, four phases are mandatory:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
plan-debate
Multi-agent debate planning pipeline for high-quality implementation plans. Uses agent separation and independent review to catch issues before code is written.
When to Use
- Tasks with 3+ implementation steps and dependencies
- Architecture decisions or system design
- Irreversible operations (database migrations, deployments, data changes)
- Any task where a wrong plan is expensive to undo
For simple tasks (1-2 steps, no dependencies), plan inline without this skill.
Pipeline Overview
| Phase | Actor | Output |
| 1. Initial Plan | planner (Sonnet) or architect (Opus) agent | Structured implementation plan |
| 2. Review | plan-reviewer (Opus) agent | Structured review with verdict |
| 3. Revision | planner + reviewer (max 2 rounds) | Revised plan if needed |
| 4. Present | orchestrator | Final plan for user approval |
| 5. Execute | orchestrator | Step-by-step implementation with task tracking |
Agent Roles
Planner (Sonnet)
- Standard feature implementation (3-5 steps)
- Explores codebase, references actual file paths
- Produces: Goal, Steps, Risks, Verification Plan, Success Criteria
Architect (Opus)
- System design, architecture decisions (5+ steps)
- Same output format as planner, deeper analysis
Plan-Reviewer (Opus)
- Independent code-aware review
- Reads source files to verify plan assumptions
- Cross-checks against project rules and error lessons
- Outputs structured review with verdict
Verdict Types
| Verdict | Meaning | Action |
| APPROVE | Plan is sound | Present to user for approval |
| REVISE | Fixable issues found | Re-plan with feedback (max 2 rounds) |
| REJECT | Fundamental problems | Ask user for guidance |
Verification Tiers
Every plan must include a Verification Plan with applicable tiers:
| Tier | Items | When Required |
| Baseline | lint + type-check + build + existing tests | Always |
| Unit tests | New/modified tests for changed logic | Logic changes |
| E2E tests | Playwright tests for user flows | Default-on (opt-out only) |
| Security | security-reviewer agent or manual audit | Auth, payment, data, APIs |
| Regression | Full test suite vs main branch | Breaking changes, refactors |
Research-only plans (no code changes) skip verification tiers entirely.
Frontend Workflow
When a plan touches UI, four phases are mandatory:
- Design —
/frontend-designskill produces wireframe + design direction - Implementation — build code following the design
- Spec generation —
/ui-spec capturegenerates spec from implemented code - Fidelity check — compare spec against original design, fix deviations
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.