Bug Hunter — Bundle

The problem

Developers often rush to apply fixes before understanding the root cause, leading to "whack-a-mole" debugging where the same bug resurfaces or new ones are introduced. Without a methodical reproduction and isolation process, hours are wasted on guesswork and unverified theories.

What it does

• Enforces a strict reproduce-isolate-fix workflow for every reported defect.
• Builds reproduction scripts or test cases before proposing any code changes.
• Tracks multiple hypotheses explicitly, documenting evidence for and against each theory.
• Isolates variables by changing only one element per experiment to ensure conclusive results.
• Validates every fix with a specific test or check to prove the bug is dead.

Why this beats prompting it yourself

Standard LLMs tend to be "agreeable" and will often hallucinate fixes or jump to conclusions based on symptoms. This skill is hard-coded to be skeptical and unhurried, refusing to guess and demanding evidence-based engineering over optimistic patching.

Use cases

• Isolating intermittent race conditions in backend services.
• Root-cause analysis of cryptic stack traces or memory leaks.
• Refactoring brittle code by identifying the actual failure point.
• Debugging complex logic errors in legacy codebases without documentation.

Known limitations

This skill will not perform production deployments, incident command, or destructive state-changing commands like force-pushes without explicit confirmation.