three-role-model
v0.13.0The 3-role development model as a Claude Code plugin: planner -> plan-review -> executor -> execution-review, each a separate subagent, never self-review. Also ships Autonomous Pipeline Mode — a Stop hook that catches stopping to ask permission for an already-approved step, plus a post-compact resume sequencer (ELI5 + 3-tier re-sequence + one approve-gate) and its doctrine. Bundles the role-enforcing hooks, the orchestration skills, a memory-search shim, the standalone doctrine, and a scaffolding command that makes the model the default for new skills, agents, hooks, and commands.
By ziyilam3999License: MIT1 GitHub starsUpdated 5 days ago
Directory evidence
- Runtimes
- Claude Code
- Parsed components
- 5 skill or MCP entries
- Source updated
- Sep 18, 2026
- Manifest status
- Canonical path parsed
The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →
Install three-role-model for Claude Code
claude plugin marketplace add IchenDEV/agent-plugin-mkt
claude plugin marketplace update agent-plugin-marketplace
claude plugin install three-role-model@agent-plugin-marketplacePaste and run these commands in a terminal with Claude Code. They add and refresh the PluginsMP catalog, then install this plugin.
The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.
Get the source manually
git clone https://github.com/ziyilam3999/three-role-modelClone the source repository, then follow its setup instructions to add the plugin to a compatible client. The repository root is the plugin root.
Plugin files
├── .claude-plugin/plugin.json├── skills/coherent-plan/SKILL.md├── skills/delegate/SKILL.md├── skills/double-critique/SKILL.md├── skills/issue-to-ship/SKILL.md└── skills/ship/SKILL.md
Included Skills5
Quick consistency review for small plans and strategy docs. Finds contradictions, fixes them, and produces a coherent final version. Use when the user says "/coherent-plan", "review this plan", "check for contradictions", or wants a fast consistency pass on a plan file (under ~150 lines). For large implementation specs (200+ lines with architecture and file specs), use /double-critique instead.
Mechanical planner-to-executor handoff clerk for the planner/executor workflow. Validates a plan file, pre-flights its binary AC against master in an isolated worktree, renders a brief from a template, delivers it via subagent spawn or mailbox send, and records run data. Use when the user says "/delegate PLAN" to hand a plan to an implementer, "/delegate review PR" to spawn a stateless reviewer on shipped work, or "/delegate gate PLAN" on the executor side to run the plan's AC against the current branch before push. The skill never decides whether to delegate (user's call), never writes code (implementer's job), never edits plan files, and never commits on behalf of the implementer. It is the mechanical clerk — the judgment overlay stays in CLAUDE.md.
Deep multi-agent critique pipeline that finds logical gaps, unsupported claims, missing edge cases, and internal contradictions in any document — then fixes them. Two independent critics review the document cold (seeing nothing about how it was made) to catch problems the author is blind to. Sweet spot is outcome-focused plans (WHAT+WHY); the doctrine guard and mid-loop drift-into-how detection keep HOW-creep out across the loop. Use when the user wants to review, critique, stress-test, or improve a PRD, SPEC, design doc, outcome plan, RFC, or any important document before it ships. Also use when the user says "critique this", "review this document", "find problems in", "stress test", "is this document solid", "what did I miss", "double critique", "run critique pipeline", "/double-critique path/to/file". Do NOT use for simple proofreading, grammar fixes, or formatting — this is a deep structural and logical review.
Orchestrator skill that takes a one-line issue statement, walks the operator through root-cause diagnosis, generates a Plan-First plan from template, dispatches the /auto-flow Stage-1 reviewer chain, gates on user approval, then triggers /delegate executor + /ship pipeline + push + pull. Use when an issue surfaces (incident, walk finding, retrospective, repeating mistake) and you want one entrypoint instead of weaving /auto-flow + /delegate + /ship + /cairn manually. Trigger phrases include "/issue-to-ship", "issue to ship", "diagnose and ship a fix for", "codify this as a plan and ship it", "we just hit X — turn it into a fix".
Full git shipping pipeline: commit, branch, push, create PR, wait for CI, self-review loop (with bug fix iterations), and merge. Use when the user says "/ship", "ship it", "ship this", "commit and merge", "push and merge", "create PR and merge", or wants to go from working changes to a merged PR in one command. Do NOT use for simple git operations like just committing or just pushing -- this is the full end-to-end pipeline.
Plugin manifests1
{
"name": "three-role-model",
"description": "The 3-role development model as a Claude Code plugin: planner -> plan-review -> executor -> execution-review, each a separate subagent, never self-review. Also ships Autonomous Pipeline Mode — a Stop hook that catches stopping to ask permission for an already-approved step, plus a post-compact resume sequencer (ELI5 + 3-tier re-sequence + one approve-gate) and its doctrine. Bundles the role-enforcing hooks, the orchestration skills, a memory-search shim, the standalone doctrine, and a scaffolding command that makes the model the default for new skills, agents, hooks, and commands.",
"version": "0.13.0",
"author": {
"name": "ziyilam3999"
},
"homepage": "https://github.com/ziyilam3999/three-role-model",
"repository": "https://github.com/ziyilam3999/three-role-model",
"license": "MIT"
}For maintainers
If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.
[three-role-model on Agent Plugins Marketplace](https://pluginsmp.com/plugins/three-role-model)