| name | reproduce-and-fix-bug |
| title | Reproduce and Fix Bug |
| description | Turn a bug report into an automated failing test that reproduces it, then make that test pass with the smallest correct fix. Use when the user reports a bug, pastes a stack trace or error message, or says "reproduce this", "write a failing test", "fix this bug", "why does X break", or "this returns the wrong result". |
| category | debug-understand |
| tools | ["read_file","glob","grep","write_file","edit_file","Bash(git diff *)","Bash(git log *)","Bash(pytest *)","Bash(npm test *)","Bash(go test *)","Bash(cargo test *)"] |
Reproduce and Fix Bug
Reproduce the bug as a failing test first, then fix it. A test that fails before and passes after is the proof — never skip it.
Step 1: Extract the reproduction
From the report, pin down:
- Exact inputs / steps and the environment.
- Expected behavior vs actual (error, wrong value, crash).
- Any stack trace — note the failing file, function, and line.
If critical details are missing, state your assumptions explicitly and proceed with the most likely case.
Step 2: Locate the code path
grep -rn the error message, symbols, and frames from the trace to find the involved functions.- Read the implicated code and its callers; form a hypothesis about the fault.
Step 3: Learn the test harness
glob for existing tests (test_*, *_test.*, *.spec.*) near the affected code.- Read one to copy conventions (fixtures, naming, assertions).
- Confirm how tests run and that the suite is green now:
pytest -q / npm test / go test ./... / cargo test.
Step 4: Write a failing test
- Add a focused test that encodes the expected behavior for the reported input.
- Run just that test and confirm it FAILS — and fails for the real reason (asserts on the bug), not a setup/import error.
- If it passes unexpectedly, your reproduction is wrong; refine inputs until it fails as reported.
Step 5: Diagnose the root cause
- Trace from the failing assertion to the origin; distinguish the root cause from its symptom.
- Avoid fixing at the surface if the defect is upstream.
Step 6: Apply the minimal fix
- Use
edit_file for the smallest change that corrects the root cause.
- Do not reformat unrelated code or bundle in extra changes.
Step 7: Verify
- Re-run the new test — it must PASS.
- Run the surrounding suite to catch regressions: full module or
pytest -q / npm test / go test ./....
- Consider adjacent edge cases (empty, null, boundary, negative) and add a case if warranted.
Step 8: Report
- Show
git diff of the fix plus the new test.
- State the root cause in one sentence and why the change resolves it.
- Note any remaining risks or follow-ups. Do not commit unless asked.
