Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
44 lines (29 loc) · 2.41 KB

CONTRIBUTING.md

File metadata and controls

44 lines (29 loc) · 2.41 KB
Raw

Contributing

Contributions that improve the orchestration protocol, add failure scenarios, or sharpen the worked examples are welcome.

What to Contribute

Good contributions:

  • Corrections to protocol steps that cause incorrect or ambiguous behavior in real sessions
  • New failure scenarios based on observed Agent Teams failures (with root cause + resolution + anti-patterns)
  • Improvements to the references/example.md that make the worked example clearer or more representative
  • Clarifications to the Teammate Prompt Template or Synthesis Protocol

Not in scope:

  • Changes that require Agent Teams features not yet available in Claude Code
  • Additions that conflict with the minimum viable team principle or exception-driven communication model

How to Contribute

  1. Open an issue first for anything beyond a small typo fix. Describe the problem and your proposed solution before writing content.

  2. Fork and branch from main.

  3. Make focused changes. Each PR should address one concern. Mixing a bug fix with a new failure scenario makes review harder.

  4. Validate against real usage. If you are changing orchestration instructions, test the updated skill in an actual Claude Code Agent Teams session. Document what you tested in the PR.

  5. Keep SKILL.md internally consistent. Every section of the skill file is read by the model during execution. Contradictions between phases, the Teammate Prompt Template, and the Synthesis Protocol cause real failures. Cross-check your changes against all other sections.

  6. Update references if needed. If your change affects observable behavior shown in references/example.md, update the example. If you are fixing a failure mode, add or update the corresponding entry in references/failure-scenarios.md.

PR Checklist

  • Change is focused on a single concern
  • SKILL.md is internally consistent after the change
  • Tested in a real Agent Teams session (or clearly marked as untested)
  • References updated if behavior changed
  • Issue linked in PR description

Style Notes

  • Use second person ("The lead calls...") for protocol instructions in SKILL.md
  • Use present tense for facts, imperative for actions
  • Anti-patterns sections use bold labels: **DO NOT:** followed by the problematic behavior in a blockquote
  • Failure scenario entries follow the structure: Situation → Root Cause → Resolution (numbered steps) → Anti-patterns