https://github.com/pcliangx/appgenesisforge
Production-ready Claude Code Agent Team template — 12 roles, 3-layer hook defense, multi-LLM (DeepSeek/Doubao/Qwen/MiniMax) + React/FastAPI/Postgres + WeChat MiniProgram. 生产级 Claude Code 多 Agent 协作模板。
https://github.com/pcliangx/appgenesisforge
agent-team agentic-ai ai-agents claude claude-code claude-code-agent deepseek doubao fastapi fullstack-template llm minimax multi-agent prompt-engineering qwen react sub-agents template wechat-miniprogram
Last synced: 15 days ago
JSON representation
Production-ready Claude Code Agent Team template — 12 roles, 3-layer hook defense, multi-LLM (DeepSeek/Doubao/Qwen/MiniMax) + React/FastAPI/Postgres + WeChat MiniProgram. 生产级 Claude Code 多 Agent 协作模板。
- Host: GitHub
- URL: https://github.com/pcliangx/appgenesisforge
- Owner: pcliangx
- License: mit
- Created: 2026-04-28T03:36:39.000Z (21 days ago)
- Default Branch: main
- Last Pushed: 2026-05-03T03:03:56.000Z (16 days ago)
- Last Synced: 2026-05-03T12:10:53.782Z (16 days ago)
- Topics: agent-team, agentic-ai, ai-agents, claude, claude-code, claude-code-agent, deepseek, doubao, fastapi, fullstack-template, llm, minimax, multi-agent, prompt-engineering, qwen, react, sub-agents, template, wechat-miniprogram
- Language: Shell
- Homepage:
- Size: 806 KB
- Stars: 4
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.en.md
- Changelog: CHANGELOG.md
- License: LICENSE
Awesome Lists containing this project
README
# App Genesis Forge · Your AI Coworker Team
> **Code the Origin, Forge the App.**
> 14 AI coworkers · one workflow · ship a feature overnight.
[](https://github.com/pcliangx/AppGenesisForge/generate)
[](https://github.com/pcliangx/AppGenesisForge/releases)
[](https://opensource.org/licenses/MIT)
[](https://claude.com/claude-code)
[](.claude/agents/)
[](.claude/skills/agf-wiring-multi-llm-sdk/)
[](.claude/standards/miniapp.md)
[](.claude/standards/security.md)
[](https://github.com/pcliangx/AppGenesisForge/stargazers)
[](https://github.com/pcliangx/AppGenesisForge/pulls)
[简体中文](README.md) · **English**
---
## 💡 What Is This?
Built on [Claude Code Agent Teams](https://code.claude.com/docs/en/agent-teams) (experimental feature; reuses sub-agent definitions in `.claude/agents/`) — not a prompt collection, not a roster of agent personalities, but a **runnable AI team scaffold with workflow, standards, and acceptance gates**, born from a real team's iteration on "can AI collaboration actually ship a feature".
- **🎯 Hard role boundaries**: 14 specialists each in their lane — PL never codes, Reviewer never edits source, QA never makes business judgment calls
- **🧠 Personality + non-negotiable rules**: war-scar vibe lines ("PTSD about schema migrations") plus rules that cannot be bypassed
- **📋 PRD → UAT full chain**: every feature passes `code review → SIT → E2E → UAT` stage gates; artifacts land in `docs/`
- **🛡️ Safe by default**: 4-layer hook defense (dangerous commands / secrets / injection / commit) out of the box
- **🇨🇳 China-native**: DeepSeek / Doubao / Qwen / MiniMax + WeChat Mini Program trio + monthly cost governance
**TL;DR**: Upgrades "AI team" from a roster of actors to a factory line that signs off, pushes back, and leaves evidence.
---
## 👋 Meet the Team
We've hired **14 AI coworkers** for you — a non-meddling product lead, a tech advisor who only shows up when it matters, six engineers who work in parallel, two reviewers whose only job is finding flaws, two QAs who refuse to sign off without evidence, plus a content writer for release notes and a growth analyst for post-launch experiments.
They don't sleep, don't race for commits, and don't bypass workflow. Every collaboration step and artifact path is written on the wall.
This isn't a prompt collection — it's an AI team with **process, standards, and acceptance gates**. v3.0.0 rolls agent names back to bare form (real-world Agent Teams routing reasons); skills + commands keep `agf-` prefix — see [`CHANGELOG.md`](CHANGELOG.md).
**Built for**: React + FastAPI + Postgres full-stack · AI Agent / RAG / multimodal inference (China-domestic LLMs: DeepSeek / Doubao / Qwen / MiniMax) · WeChat Mini Programs · full PRD-to-UAT delivery loop.
**Not for**: native iOS/Android · desktop GUI · model training · Java/Go/Rust mainline (would need `tech-lead` to rewrite ADR-000).
---
## ⏱ Day 1
> **Prerequisite**: Claude Code ≥ v2.1.32 with Agent Teams enabled (`.claude/settings.json` ships the `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` flag; `init-team.sh` validates it). New to Claude Code? See the [official setup guide](https://code.claude.com/docs/en/setup).
### Option 1 — Scaffold a new project (recommended)
```bash
gh repo create my-app --template pcliangx/AppGenesisForge --private
cd my-app
bash init-team.sh
```
`init-team.sh` runs 10 verification checks (Claude Code version, hook regression, JSON validity, required files, git status) and installs the git pre-commit hook automatically.
### Option 2 — Embed into an existing repo
Copy `.claude/`, `CLAUDE.example.md` (rename to `CLAUDE.md`), `init-team.sh`, `evals/`, and the `docs/` templates. Then run `bash init-team.sh`. **Do not copy this repo's own `CLAUDE.md`** — that's the template's project description, not the template itself.
### Option 3 — Run your first chain
```text
/agf-team-start
```
Full 10-step Day-1 checklist + common pitfalls: [`.claude/FIRST_RUN.md`](./.claude/FIRST_RUN.md).
---
## 🏢 The 14 Coworkers
```mermaid
flowchart TD
U([User])
PL[product-lead
orchestrator · opus]
TL[tech-lead
conditional · opus]
EXE[Execution Layer
6 engineers]
CR[code-reviewer
review-only]
QA[qa-engineer
SIT → E2E → UAT]
U -->|requirement| PL
PL -.->|architecture Q| TL
TL -.->|ADR / risk| PL
PL -->|tasks + AC| EXE
EXE -->|completion| PL
PL -->|review| CR
CR -->|verdict| PL
PL -->|tests| QA
QA -->|report| PL
PL -->|delivery| U
style PL fill:#f97316,color:#fff
style TL fill:#3b82f6,color:#fff
style EXE fill:#22c55e,color:#fff
style CR fill:#eab308,color:#fff
style QA fill:#ef4444,color:#fff
```
| # | Coworker | Role | When to call |
|---|---|---|---|
| 🟠 #001 | [`product-lead`](.claude/agents/product-lead.md) | Sole orchestrator: PRD, task assignment, final sign-off | Any complete feature; multi-role coordination; UAT |
| 🔵 #002 | [`tech-lead`](.claude/agents/tech-lead.md) | Conditional advisor: ADR, tech-stack baseline | Missing baseline / new tech selection / risk escalation |
| 🟢 #003 | [`frontend-dev`](.claude/agents/frontend-dev.md) | React + Vite, API integration | UI components, styling, frontend bugs |
| 🟢 #004 | [`backend-dev`](.claude/agents/backend-dev.md) | FastAPI / SQLAlchemy / Alembic | REST APIs, DB migrations, auth middleware |
| 🟢 #005 | [`ai-agent-dev`](.claude/agents/ai-agent-dev.md) | LLM integration, prompt engineering | RAG, tool calling, guardrails |
| 🟢 #006 | [`ml-engineer`](.claude/agents/ml-engineer.md) | Multimodal inference, image/video pipelines | Doubao 2.0 Pro / Kling / MiniMax Video |
| 🟢 #007 | [`uiux-designer`](.claude/agents/uiux-designer.md) | UI/UX + standalone HTML prototypes | Layouts, clickable prototype reviews |
| 🟢 #008 | [`miniapp-dev`](.claude/agents/miniapp-dev.md) | WeChat Mini Programs (native / Taro fallback) | Pages, wx.* APIs, network, unit tests |
| 🟡 #009 | [`code-reviewer`](.claude/agents/code-reviewer.md) | Review-only: code, security, quality | PR review, vulnerability audits |
| 🟡 #010 | [`miniapp-code-reviewer`](.claude/agents/miniapp-code-reviewer.md) | Mini-program review (compliance + bundle size) | wx.* APIs, setData perf, sub-package size |
| 🔴 #011 | [`qa-engineer`](.claude/agents/qa-engineer.md) | SIT / E2E / UAT full chain | System integration tests, AC validation |
| 🔴 #012 | [`miniapp-qa-engineer`](.claude/agents/miniapp-qa-engineer.md) | Mini-program simulator + real-device testing | Real-device regression, privacy compliance |
| 🟪 #013 | [`content-writer`](.claude/agents/content-writer.md) | Release notes / blog / user cases | Post-UAT writeups, launch posts |
| 🟪 #014 | [`growth-analyst`](.claude/agents/growth-analyst.md) | North-star / OMTM / A/B experiments | Metric definition, experiment design, post-launch validation |
Full capability matrix & Mermaid map: [`docs/team-capability-map.md`](docs/team-capability-map.md). Delivery workflow & glossary: [`docs/product-workflow.md`](docs/product-workflow.md).
---
## 🤝 How They Work Together
A typical Monday: you run `/agf-team-start add team-invite flow`.
1. **#001** writes the PRD and posts acceptance criteria
2. **#007** ships an HTML prototype to `docs/design/[feature]/index.html`
3. **#003** + **#004** implement frontend + backend in parallel (separate worktrees)
4. **#009** reviews the PR — security, quality, conventions
5. **#011** runs SIT/E2E/UAT and writes the report to `docs/qa/`
6. **#001** signs off and closes the loop
One slash command produces a full feature. PRD, design, code, review, and test reports all archived under `docs/` per [`.claude/rules/repo-layout.md`](.claude/rules/repo-layout.md).
Other workflows (multi-LLM provider swap, WeChat mini-program submission, monthly cost audit, weekly ADR drift check) follow the same pattern with different rosters — see the [Chinese README](README.md#-他们怎么协作--how-they-work-together) for the full set.
---
## 🛠️ Tooling at a Glance
- **7 hooks** — 4 hard defenses (`block-dangerous-bash`, `scan-secrets`, `sanitize-tool-output`, `scan-commit`) + 3 workflow nudges (`teammate-keepalive`, `plan-sync-check`, `skill-suggester`)
- **5 skills** — `agf-writing-prd`, `agf-writing-adr`, `agf-running-sit-tests`, `agf-writing-sit-report`, `agf-wiring-multi-llm-sdk`
- **4 slash commands** — `/agf-team-start`, `/agf-usage`, `/agf-sit`, `/agf-uat`
- **12 standards** — roles, workflow, AC lifecycle, testing, security, observability, cost, documentation, coding, superpowers, miniapp, versioning
Full standards: [`.claude/standards/`](.claude/standards/) · Chinese deep-dive: [README.md](README.md).
---
## ♻️ Reuse · License · Roadmap
**Reuse**: new project → `gh repo create --template`. Existing repo → copy `.claude/` + `init-team.sh`, run init. Legacy codebase → let `tech-lead` scan the repo to derive ADR-000 from current state, never green-field selection.
**Roadmap**: ✅ v2.0.0 namespace migration · ✅ 4 China-domestic LLMs · ✅ WeChat MiniProgram trio · ⏳ more skills (test cases, commit messages, release notes) · ⏳ CI integration of `init-team.sh` · ⏳ video walkthrough.
**License**: MIT — free for commercial use, attribution appreciated. Structural inspiration from [The Agency](https://github.com/msitarzewski/agency-agents).
---
**14 AI coworkers · one workflow · ship a feature overnight**
[⭐ Star](https://github.com/pcliangx/AppGenesisForge) · [🍴 Fork](https://github.com/pcliangx/AppGenesisForge/fork) · [🐛 Issue](https://github.com/pcliangx/AppGenesisForge/issues) · [📒 CHANGELOG](CHANGELOG.md) · [🇨🇳 简体中文](README.md)