autocode-flow

Automate your development pipeline with a single command for project-specific workflows.

autocode-flow

One command, fully autonomous development. Describe a feature, get working tested code.

/autocode-flow:autocode add user login & registration
PLAN ──► TDD ──► CODE ──► TEST ──► VERIFY ──► DONE
 (all automatic, zero human intervention)

How It Works

Step 1: Setup (once per project)

/autocode-flow:autocode-new

Scans your project, asks a few questions, generates a project-specific pipeline config.

Step 2: Build features (every time)

/autocode-flow:autocode add user login & registration
/autocode-flow:autocode prd @docs/user-auth-prd.md
/autocode-flow:autocode Fix the race condition in order processing

The pipeline runs fully automatically:

PhaseWhat Happens
PLANDecomposes requirement into ordered atomic steps
TDDWrites all test files first (RED phase)
CODEImplements minimum code per step until GREEN
TESTRuns full suite, loops until coverage target met
VERIFYLints, self-reviews, auto-fixes quality issues
E2EWrites & runs E2E tests (if enabled)
DONEReports results, suggests git commit

No human intervention between phases. It plans, codes, tests, and verifies in one shot.

Installation

Option 1: From GitHub

claude plugin marketplace add charles001-wong/autocode-flow
claude plugin install autocode-flow@autocode-flow

Option 2: From local directory

claude plugin install --path ~/autocode-flow

Option 3: Manual

mkdir -p ~/.claude/plugins/cache/autocode-flow/autocode-flow/1.0.0
cp -r . ~/.claude/plugins/cache/autocode-flow/autocode-flow/1.0.0/

Add to ~/.claude/settings.json:

{
  "enabledPlugins": {
    "autocode-flow@autocode-flow": true
  }
}

What Gets Generated

After /autocode-flow:autocode-new, your project gets:

your-project/
├── .claude/
│   ├── commands/
│   │   └── autocode.md          ← /project:autocode <feature> entry point
│   └── agents/
│       ├── planner.md           ← Task decomposition
│       ├── tdd-guide.md         ← Test-first enforcement
│       ├── code-reviewer.md     ← Quality verification
│       └── e2e-runner.md        ← E2E tests (if enabled)
├── .cursor/
│   ├── rules/
│   │   ├── autocode-workflow.mdc
│   │   ├── coding-style.mdc
│   │   └── testing.mdc
│   └── skills/
│       └── autocode/
│           ├── SKILL.md
│           └── config.json      ← Your pipeline settings
└── docs/autocode/pipeline.md

All files are project-specific — generated with your actual language, framework, test commands, and linter.

Supported Stacks

LanguageFrameworksTest Tools
Gogo-zero, Gin, Echo, Fibergo test, testify
TypeScriptNext.js, Vite, Angularvitest, jest, Playwright
JavaScriptExpress, React, Vuejest, vitest, Cypress
PythonFastAPI, Django, Flaskpytest, hypothesis
RustActix, Axumcargo test

Example: Go + go-zero

> /autocode-flow:autocode-new
📋 Detected: Go / go-zero / golangci-lint / GitHub Actions
... (answers a few questions) ...
✅ Pipeline generated!

> /autocode-flow:autocode 新增用户注册功能

═══ Phase 1: PLAN ═══
  4 steps: model → logic → handler → integration test

═══ Phase 2: TDD ═══
  Writing tests... 12 test cases across 4 files
  Running: go test ./... → FAIL (expected)

═══ Phase 3: CODE ═══
  Step 1/4: user model ✅
  Step 2/4: register logic ✅
  Step 3/4: register handler ✅
  Step 4/4: integration wiring ✅

═══ Phase 4: TEST ═══
  go test ./... → PASS (12/12)
  Coverage: 87% (target: 80%) ✅

═══ Phase 5: VERIFY ═══
  golangci-lint run → clean ✅
  Self-review → no issues ✅

═══════════════════════════════════════
  autocode complete: 用户注册功能
  Tests: 12 cases, Coverage: 87%
  Files: 8 created
  Ready: git add -A && git commit -m "feat: add user registration"
═══════════════════════════════════════

Architecture

autocode-flow (plugin)
├── /autocode-flow:autocode-new ── scan → interview → generate config
├── /autocode-flow:autocode ────── load config → plan → tdd → code → test → verify
├── agents/ ───────────────────── scanner, interviewer, generator
├── templates/ ────────────────── project-specific file templates
└── scripts/ ──────────────────── scan-project.sh (tech stack detection)

License

MIT