Chapka

An insulated and opinionated starter kit for participating in the open source software community in an AI-first software development life cycle.

Chapka provides ready-to-copy scaffolding for AI coding agents so that contributors using these tools can navigate disclosure requirements, licensing concerns, community norms, and quality expectations across diverse open source projects.

The problem

Open source communities are rapidly establishing norms for AI-assisted contributions. Policies range from permissive-with-guardrails (Linux kernel, Fedora) to restrictive (attrs, Gentoo). A contributor using AI tools today must understand attribution formats, licensing risks, DCO/CLA constraints, and anti-spam expectations -- and these vary by project. Getting it wrong wastes maintainer time and damages trust.

Chapka encodes the common ground across these policies into agent-consumable configuration files, so your AI tool behaves responsibly by default.

What's in the box

chapka/
  ASSESSMENT_Claude.md       # Landscape analysis: how communities approach AI contributions
  ASSESSMENT_Gemini.md       # Same analysis, independently synthesized
  POLICY_LIST.md             # Raw link collection of AI policies and discussions
  PROMPT.md                  # The prompt used to generate the assessments

  cursor/                    # Scaffolding for Cursor users
    .cursor/rules/           # .mdc rule files (core behavior + attribution)
    AGENTS.md                # Agent context manifest
    CHECKLIST.md             # Pre-PR checklist
    RESOURCES.md             # Policy links + Cursor docs
    templates/               # PR body and commit footer templates

  claude/                    # Scaffolding for Claude Code users
    CLAUDE.md                # Template project context (copy to repo root)
    .claude/commands/        # Custom slash commands
    .claude/skills/          # Specialized agent skill definitions
    settings.json            # Example hooks and permission guards
    CHECKLIST.md             # Pre-PR checklist
    RESOURCES.md             # Policy links + Claude Code docs
    templates/               # PR body and commit footer templates

Quick start

For Claude Code users

1. Copy claude/CLAUDE.md to the root of the project you're contributing to.
2. Copy claude/.claude/commands/ to .claude/commands/ in that project.
3. Copy claude/.claude/skills/ to .claude/skills/ (or selectively promote skills to commands).
4. Merge relevant hooks from claude/settings.json into .claude/settings.json.
5. Add project-specific details (build commands, test paths, architecture) to the CLAUDE.md you copied.

Available slash commands:

• /project:pre-pr-check -- audit staged changes against the OSS checklist
• /project:commit-assisted -- draft a commit with proper Assisted-by: trailer
• /project:check-upstream-policy -- search the repo for AI contribution policies

Available skills (copy to .claude/commands/ to use as slash commands):

• code-review -- first-pass review for correctness, style, and AI-specific concerns
• test-gen -- generate tests matching the project's framework and conventions
• docs-sync -- draft doc updates when code changes alter APIs or behavior
• license-audit -- check AI-generated code for provenance and license risk
• security-scan -- audit AI output for common vulnerability patterns
• upstream-onboard -- research a new project's policies and conventions
• pr-prepare -- full compliance sweep before opening a pull request

For Cursor users

1. Copy cursor/.cursor/rules/ to the target project.
2. Copy cursor/AGENTS.md to the project root.
3. Add project-specific context to AGENTS.md.

See cursor/README.md for details.

Core principles

These are consistent across both tool scaffolds and reflect the consensus found in the ASSESSMENT analyses:

• Humans are accountable. The contributor owns every line. "An LLM wrote it" is never a defense.
• AI is not an author. No Co-authored-by: for tools. Use Assisted-by: trailers for disclosure.
• Existing standards apply. AI-assisted contributions must meet the same quality, licensing, and review bar as any other.
• Upstream wins. The target project's policies override Chapka defaults. Always check first.
• Honest disclosure. When AI materially shaped the contribution, say so. Trivial assistance (grammar, spelling) is generally exempt.

What the assessments cover

The ASSESSMENT files synthesize policies from the Linux kernel, Fedora, attrs, Zulip, Pulp, Avocado, Node.js, Django, OpenSSL, and others. They cover:

Theme	Key question
Verification and quality	Can you explain and defend every line?
Attribution and disclosure	When and how to declare AI usage?
Licensing and copyright	Does AI output risk license contamination?
Contributor accountability	Who is legally responsible?
Spam and abuse prevention	Is this a valuable, scoped contribution?
Enforceability	How do projects actually enforce these norms?

Adding support for another tool

Chapka is designed to grow. To add scaffolding for a new AI coding tool:

1. Create a directory named after the tool (e.g., windsurf/, copilot/).
2. Translate the core principles and attribution rules into that tool's native configuration format.
3. Add a README explaining how the tool's features map to the ASSESSMENT themes.
4. Include a checklist, resource list, and templates following the same structure.

Related resources

• CHAOSS WG AI Alignment -- curated list of 20+ project and foundation AI policies
• ASF Generative Tooling Guidance
• Linux Foundation Generative AI Policy
• Fedora AI-Assisted Contributions Policy

Part of Chapeaux

Chapka is part of the Chapeaux family of projects.

License

See the project license file for terms.