# Bug Fix

Write failing test → fix → verify

## Agents

- test-writer
- code-fixer
- test-runner

## Execution

Run the agents in sequence (or parallel where noted). Collect all outputs in
`.claude/audits/`.