openclaw
/
maintainers
mirror of https://github.com/openclaw/maintainers.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
onboarding and skills for maintainers
10 Commits 2 Branches 0 Tags 105 KiB
Markdown 100%
Go to file
Josh Lehman 9178dba968
fix: instruct changelog entries to append at end of section (#3) 
Codex agents were inserting changelog entries near the top of
sections, causing unnecessary conflicts. Add explicit guidance
to append to the end of the relevant section.
 		2026-03-06 01:49:22 +00:00
.agents/skills fix: instruct changelog entries to append at end of section (#3) 2026-03-06 01:49:22 +00:00
README.md
…

README.md

New Maintainer Guide

A survival guide for new OpenClaw maintainers. Covers setup, security, first PRs, processes, and where to find help. For deeper reference, check the full knowledge base or ask in #maintainers.

Setup

# Clone and build
git clone https://github.com/openclaw/openclaw.git
cd openclaw
pnpm install
pnpm build            # Build (instant with tsdown)
pnpm check            # Combined lint + format + types
pnpm test             # Full test suite (~87s)

# Install maintainer skills (PR workflow tools)
npx skills add https://github.com/openclaw/maintainers

# Install OpenClaw itself
curl -fsSL https://openclaw.ai/install.sh | bash
openclaw doctor       # Run diagnostics

Security Hardening

You are now a target for phishing, social engineering, and supply chain attacks. Harden your accounts immediately.

  • Enable 2FA on GitHub using an authenticator app
  • Remove SMS as a 2FA backup (SIM swap risk)
  • Set up GPG-signed git commits
  • Get a YubiKey/FIDO2 hardware key (strongly recommended)
  • Use a password manager (1Password/Bitwarden) for all passwords
  • Rotate all API keys and tokens
  • Block random Discord/LinkedIn/Instagram friend requests
  • Never click unverified cal.com scheduling links
  • Don't run OpenClaw with a personal identity that has access to sensitive systems
  • Verify identity through a separate channel before any sensitive action

Getting Started

Orient and Observe

  • Follow threads in #maintainers (the main discussion channel)
  • Read pinned messages in #maintainers
  • Read the PR workflow: .agents/skills/PR_WORKFLOW.md
  • Browse small PRs: filter by is:open is:pr label:size:xs
  • Pick one and read it thoroughly — don't take action yet
  • Run /reviewpr on it to see the AI review output
  • Read 3-5 recently merged PRs to learn the pattern

Your First PRs

  • Assign yourself to a size:xs PR ("if you lick it, it's yours")
  • Review with AI: /reviewpr in Codex
  • Evaluate: What's the actual problem? Is this the most optimal fix?
  • Rework if needed — rewriting contributor code is normal and expected
  • Run gates: pnpm lint && pnpm build && pnpm test
  • Update CHANGELOG.md with the PR number and thank the contributor
  • Merge using /landpr or /mergepr

Important: Treat PRs as problem descriptions, not finished code. Fixing up code before merging is standard practice.

Think in Clusters, Not Individual PRs

At OpenClaw's scale, you'll rarely deal with a single isolated issue. When you spot something interesting, use Codex to search for similar or duplicate entries — you'll usually find a whole cluster: multiple PRs attempting the same fix, related issues from different angles, and various approaches people have tried.

The workflow: find the cluster → have Codex review the whole thing → identify the best PR to base a fix from → merge that one → close the duplicates. A single good merge often resolves 5+ related issues and PRs at once. This is far more effective than processing PRs one at a time.

Build Your Rhythm

  • Run the gateway locally: pnpm gateway:watch (requires pnpm build first)
  • Explore key source files: src/gateway/, src/auto-reply/envelope.ts, src/config/, src/agents/
  • Graduate to size:s PRs
  • Begin closing obvious junk PRs politely ("Thank you for your kind PR")

PR Filters

Filter Use
is:open is:pr label:size:xs Smallest, safest PRs to start with
is:open is:pr label:trusted-contributor Higher-quality submissions
is:open is:pr label:maintainer Other maintainer PRs (review these)

The 3-Step Workflow

Review → Prepare → Merge. See PR_WORKFLOW.md for the full process.

  1. Review: Run /reviewpr. Evaluate the problem, implementation, security impact.
  2. Prepare: Rebase, fix code, run all gates, push changes.
  3. Merge: Squash-merge, update CHANGELOG, close the PR.

Auto-Close These

  • Skills PRs — redirect to ClawHub (use clawdhub label)
  • PRs > 5K LOC — auto-closed per policy
  • @ts-nocheck or lint disabling — never acceptable
  • Rename/rebrand spam — close politely

Volume and Triage

PR and issue volume is high. Triage bots handle first-pass filtering so you can focus on promising PRs. Releases happen roughly daily. The flow: cut beta → maintainers test → fix regressions → ship stable.

Do's and Don'ts

Do

  • Run pnpm lint && pnpm build && pnpm test before every merge
  • Always rebase before merging (stale PRs break main)
  • Thank real humans in CHANGELOG entries
  • Announce in #maintainers before breaking changes
  • Show, don't tell — demos get quick approvals
  • Ask "does this exist somewhere already?" before building

Don't

  • Push untested code to main
  • Merge your own non-trivial PRs without review
  • Use bun (incompatible with upstream deps)
  • Update Carbon dependency (Shadow controls it)
  • Send AI-generated text to external maintainers' repos
  • Share anything from #maintainers externally
  • Use LLMs to write GitHub comments pretending to be human
  • Put date/time in system prompts per-message (breaks token cache, 10x cost)
  • Add @ts-nocheck or disable lint rules

Who to Ask

For team contacts, area ownership, and subsystem maintainers, see the maintainer table in CONTRIBUTING.md. For Discord handles and internal communication norms, check the pinned messages in #maintainers.

Key Files

File Why It Matters
CONTRIBUTING.md Maintainer list, subsystem ownership, PR process
VISION.md Priorities and what NOT to merge
AGENTS.md AI agent behavior guidelines
SECURITY.md Vulnerability reporting
CHANGELOG.md Updated with every merge
.agents/skills/PR_WORKFLOW.md The 3-step PR pipeline

Quick Commands

pnpm build             # Build
pnpm check             # Lint + format + types (fast)
pnpm test              # Full test suite (~87s)
pnpm tsgo              # Type check only (10x faster)
pnpm gateway:watch     # Run gateway locally (build first!)
pnpm format            # Format code
openclaw update        # Update OpenClaw
openclaw doctor --fix  # Diagnose and fix issues

Resources

Resource URL
Main repo https://github.com/openclaw/openclaw
Docs https://docs.openclaw.ai
ClawHub (skills marketplace) https://clawhub.ai
Security/trust https://trust.openclaw.ai
Maintainer skills repo https://github.com/openclaw/maintainers
Discord discord.gg/openclaw
Install script curl -fsSL https://openclaw.ai/install.sh | bash
Beta install curl -fsSL https://openclaw.ai/install.sh | bash -s -- --beta
Security reports security@openclaw.ai
Contributing/apply contributing@openclaw.ai

Once You're Rolling

  • Pick a subsystem — review the ownership table in CONTRIBUTING.md, announce your focus in #maintainers
  • Watch a release — observe the beta → stable flow in #maintainers
  • Make an original contribution — pick a bug from post-release issues, submit a proper PR

The project runs on trust. Voluntary, fun-first, informal. Ship fast, iterate, and "code > talk."