contribbot
/kənˈtrɪbɒt/ — contrib + bot, the "b" in "contrib" merges with "bot"
中文 | English
Open source collaboration assistant. Helps developers efficiently maintain and contribute to open source projects.
MCP tools + skills — todo management, upstream tracking, issue/PR workflows, multi-project oversight.
Prerequisites
Install
Claude Code
# Step 1: Add marketplace (first time only)
claude plugin marketplace add https://github.com/darkingtail/contribbot
# Step 2: Install
claude plugin install contribbot
This installs skills + MCP server (contribbot-mcp). Skills provide guided workflows, MCP server provides the tools.
Other Platforms
contribbot's MCP server works with any MCP-compatible tool. See Other Platforms Setup for Claude Desktop, Gemini CLI, Codex CLI, Cursor, Windsurf, etc.
What contribbot does for you
Most AI coding tools can read GitHub issues and create PRs. contribbot goes further — it tracks what you're working on, what changed upstream, and who's doing what across multi-maintainer repos.
vs GitHub CLI alone
| gh CLI | contribbot |
|---|
| Read issues/PRs | ✅ | ✅ |
| Create issues/PRs | ✅ | ✅ + auto-link to local todos |
| Track personal tasks | ❌ | ✅ todo lifecycle with implementation records |
| Track upstream changes | ❌ | ✅ commit-level tracking with triage |
| Multi-maintainer coordination | ❌ | ✅ claim work items, comment to GitHub |
| Fork alignment | ❌ | ✅ sync fork + cherry-pick decisions |
| Cross-stack tracking | ❌ | ✅ track React → Vue feature parity |
| Project knowledge | ❌ | ✅ persistent knowledge per repo |
Skills
Skills are guided workflows that orchestrate MCP tools. In Claude Code, trigger them by name or natural language.
| Skill | Description |
|---|
contribbot:project-onboard | New project setup — detect fork/upstream, init config, first sync |
contribbot:daily-sync | Daily check — sync fork, fetch upstream commits, skip noise, triage |
contribbot:start-task | Start working — pick todo, activate, LLM generates implementation plan |
contribbot:todo | Todo lifecycle — add, activate, claim, update, done, archive, compact |
contribbot:issue | Issue management — list, detail, create, close, comment |
contribbot:pr | PR management — list, summary, create, update, review, reply |
contribbot:pre-submit | Pre-merge check — PR review, CI status, security alerts |
contribbot:weekly-review | Weekly retrospective — contribution stats, progress, cleanup |
contribbot:fork-triage | Fork cherry-pick decisions for downstream consumers |
contribbot:dashboard | Project overview — single or cross-project |
Project Modes
contribbot auto-detects your project's relationship with upstream repos and adapts its workflow accordingly.
| Mode | Condition | What It Enables |
|---|
| none | No fork, no upstream | Issue/PR/todo management |
| fork | Has fork source | Fork sync + cherry-pick decisions |
| upstream | Has external upstream | Cross-stack commit tracking |
| fork+upstream | Both | Fork sync + cross-stack tracking |
Run /contribbot:project-onboard to auto-detect and configure.
Why fork data is stored under parent repo
When you work on darkingtail/plane (a fork of makeplane/plane), contribbot stores data under ~/.contribbot/makeplane/plane/ — the parent repo path.
Why: multiple people may fork the same repo. Storing under the parent ensures everyone's local data aligns to the same canonical repo, and sync_fork / upstream_daily always know which repo is the source of truth.
Your fork is recorded in config.yaml as the fork field:
# ~/.contribbot/makeplane/plane/config.yaml
role: admin
org: null
fork: darkingtail/plane # your fork
upstream: null
Three-layer capability
