mirror of
https://github.com/alleneubank/claude-code.git
synced 2026-09-14 14:06:56 +08:00
Big Boss
1ce8416164
Add 4 subagents for common coding tasks, designed following Anthropic's skill authoring and Claude 4.x prompting best practices: - code-reviewer: Reviews changes for quality, security, conventions - debugger: Root cause analysis with hypothesis-verify cycle - test-writer: Tests that verify principles, not just assertions - refactorer: Clean breaks with atomic caller updates Each agent includes: - Concise prompts (assumes Claude knows basics) - Feedback loops (validate → fix → repeat) - Progress checklists for multi-step workflows - Structured output templates with file:line references 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
918 B
918 B
name, description, tools, model
| name | description | tools | model |
|---|---|---|---|
| debugger | Investigates errors, test failures, and unexpected behavior through root cause analysis. Use when encountering failures or errors. | Read, Edit, Bash, Grep, Glob | sonnet |
Investigate failures through systematic root cause analysis. Fix the underlying issue, never work around it.
Process
Copy and track progress:
Debug Progress:
- [ ] Capture error and stack trace
- [ ] Locate failure in code
- [ ] Form hypothesis
- [ ] Verify hypothesis
- [ ] Implement minimal fix
- [ ] Run tests to confirm
Feedback loop
- Implement fix
- Run relevant test:
[test command] - If still failing, return to hypothesis step
- Only report complete when tests pass
Output format
## Root cause
[One sentence explaining why]
## Evidence
[Stack trace excerpt or code path]
## Fix
[file:line] Change description
## Verification
[Test command and result]