Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
39 lines (26 loc) · 1.66 KB

CONTRIBUTING.md

File metadata and controls

39 lines (26 loc) · 1.66 KB

Contributing to deepresearch

How to contribute

  1. Fork this repository
  2. Clone your fork to ~/.claude/skills/deepresearch
  3. Run ./setup to install the skills locally
  4. Edit the SKILL.md files in any skill directory
  5. Test your changes by running the skill in Claude Code
  6. Submit a pull request

What to contribute

Skill improvements: Edit the SKILL.md files to improve instructions, add edge case handling, or fix bugs in the research workflow.

Eval criteria: Add domain-specific evaluation criteria to examples/market-landscape/.research/evals.md or improve the default criteria generated by /dr-new. Good eval criteria are binary (yes/no), objective, and catch real quality issues.

Agent improvements: Improve the researcher or evaluator subagent instructions in agents/. Changes should be concrete and measurable — "search for X using Y" not "be more thorough."

Examples: Add example research projects to examples/ showing different research domains and configurations.

No build step

There is no build, compile, or package step. SKILL.md files are plain markdown read directly by Claude Code. Edit, save, test.

Testing

After making changes:

  1. Run ./setup to update symlinks
  2. Open a new Claude Code session
  3. Run the modified skill and verify it works
  4. Check that other skills still reference the correct file paths

Style

  • Keep instructions concrete and actionable
  • Avoid vague guidance ("be thorough") in favor of specific steps ("search 3 queries")
  • SKILL.md files should be self-contained — don't assume context from other files
  • Use the same sentinel values everywhere: NOT_FOUND, UNVERIFIED, CONFLICTING, INACCESSIBLE