meyverick Bug Resolution Reporting
This skill enforces a strict, five-step diagnostic triage workflow for finding, fixing, and reporting bugs. It prevents aimless guessing and ensures defects are permanently eliminated.
When to Use
- Use when investigating unexpected behavior or failing tests.
- Use when builds break or production errors occur.
- NOT for planning new features or writing initial specs.
Core Process
Phase 1: Reproduce
You cannot fix what you cannot consistently reproduce.
- Identify the exact inputs, environment, and steps required to trigger the bug.
- Gather logs and trace the
request_id.
Phase 2: Localize
Use backward slicing to isolate the exact line of code or module causing the error.
- Start at the symptom (the error log or UI bug) and trace the execution path backward.
Phase 3: Reduce
Simplify the inputs until you have the absolute minimal reproduction case.
- Strip away unrelated systems, frameworks, and network calls.
Phase 4: Fix
Apply the code correction strictly targeting the localized failure point.
- Do not refactor unrelated code during a bug fix.
Phase 5: Guard
Prevent the bug from ever returning.
- Write a regression test targeting the specific minimal reproduction case.
Common Rationalizations
| Rationalization | Reality |
|---|---|
| "I think I know what it is, I'll just change this line." | Guessing leads to regressions. You must successfully reproduce and localize the error first. |
| "I fixed it, but I don't know how to write a test for it." | If you cannot write a test for the bug, you have not sufficiently isolated it. |
| "I'll refactor this whole file while I fix the bug." | Conflating bug fixes with refactoring makes PRs impossible to safely review and rollback. |
Red Flags
- Modifying source code before producing a failing test case or verifying logs.
- PRs for bug fixes that lack accompanying regression tests.
Verification
Before finalizing the bug resolution report, verify:
- The exact root cause is documented.
- A minimal reproducible test case was created and failed initially.
- The fix was applied cleanly without unrelated refactoring.
- The regression test now passes consistently.