robo-co-op/robobuilder-standard
Developer toolsこのページは AI により自動翻訳されています。技術的な詳細は原文をご確認ください。 原文を見る
Robo Co-op standard Claude Code dev plugin — 40 skills, 9 agents, 6 hooks, bidirectional Notion Bootcamp link
概要
Primary distribution is a Claude Code plugin. The same skill source can also be exported as an OpenClaw/Codex skill pack. Standard covers the inner loop (L1 agent loop + L2 verification: tdd / diagnose / review). Pro adds the outer loop (L3 event-driven + L4 self-improving loops), and above it L5 — a graph of loops, where reliability lives in the edges: who watches whom, who can veto whom. Lite is the on-ramp to Standard. The three editions are the — how work gets done. What must be true before work is allowed to move forward is defined once, for every Robo Co-op repository, in . Every project travels the same spine, and each state has to leave evidence behind before the next one may begin: Four gates block the transitions that matter: Definition of Ready, Merge Gate, Production Gate, Definition of Done. A repository declares which OS version it follows by copying robobuilder-os/templates/AGENTS.md into its root. Coding standards stay here in CLAUDE.md.
README
以下は原文 README の抜粋です。原文を見る。
robobuilder (Standard edition)
Robo Co-op standard development skill system.
Primary distribution is a Claude Code plugin. The same skill source can also be exported as an OpenClaw/Codex skill pack.
Three editions
| Edition | Repo | Who it’s for | Contents |
|---|---|---|---|
| Lite | robobuilder-lite | Beginners — learn the workflow with 4 commands | 4 merged mega-skills (plan / build / improve / ship) + the same hooks and review agents |
| Standard (this repo) | robobuilder-standard | Daily development | 41 skills / 9 agents / 6 hooks / 3 playbooks |
| Pro | robobuilder-pro | Loop & Graph Engineering — autonomous agent loops | Add-on installed alongside Standard: design, gate, audit, and compound loops built on native /goal, /loop, /batch, and Routines — then wire them into a graph with typed edges, arbitration, and champion-challenger promotion |
Standard covers the inner loop (L1 agent loop + L2 verification: tdd / diagnose / review). Pro adds the outer loop (L3 event-driven + L4 self-improving loops), and above it L5 — a graph of loops, where reliability lives in the edges: who watches whom, who can veto whom. Lite is the on-ramp to Standard.
The operating system above the editions
The three editions are the toolbox — how work gets done. What must be true before work is allowed to move forward is defined once, for every Robo Co-op repository, in robobuilder-os.
Robo Builder = HOW. Robo Builder OS = MUST / WHEN / DONE.
Every project travels the same spine, and each state has to leave evidence behind before the next one may begin:
DEFINE -> DESIGN -> ISSUE -> BRANCH -> TEST -> BUILD -> REVIEW -> STAGE -> DEPLOY -> VERIFY -> OPERATE
| This repo’s workflow | OS state |
|---|---|
| Foundation / Investigate | DEFINE |
| Design, Prototype | DESIGN |
| Plan → issues | ISSUE |
| Implement (TDD) | BRANCH, TEST, BUILD |
| Refactor | BUILD |
Review (/diff-review, /cross-review) |
REVIEW |
| Ship | STAGE, DEPLOY, VERIFY |
Four gates block the transitions that matter: Definition of Ready, Merge Gate, Production Gate, Definition of Done.
A repository declares which OS version it follows by copying
robobuilder-os/templates/AGENTS.md into its root. Coding standards stay here in
CLAUDE.md.baseline; lifecycle rules and gates live in the OS. If a rule is about taste, it
belongs in this repo — if it decides whether a change may advance, it belongs in the OS.
One install gives every team member:
- 41 curated dev skills across 6 phases (Investigate → Design → Prototype → Implement → Refactor → Review → Ship)
- 9 phase-specific subagents
- 3 multi-skill playbooks
- An onboarding wizard, a CLAUDE.md tuning wizard, and a semver upgrade flow
推奨ツール
別のキーワードを試すか、フィルタを外してください。
インストール
npx skillfish add robo-co-op/robobuilder-standard