メニュー
Use when you hit any bug, test failure, or unexpected behavior - find the root cause before proposing a fix.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use to build an approved detailed spec task by task in the current session - tests required, single-agent by default, subagents only for large or parallel tasks.
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores intent, requirements, and design before implementation.
Use when implementation is complete and verified, to integrate the work - merge, PR, or cleanup - after a final evidence check.
Use after a design is approved in brainstorming, to turn it into one layered detailed spec (acceptance criteria, tasks, and how to verify) before touching code.
Use when starting any conversation - establishes how to find and use decentpowers skills and the workflow they form.
Use before claiming work is complete, fixed, or passing - runs the deterministic gate (tests, build, lint, types, acceptance) and one light judgment review, with evidence before any claim.
| name | debugging |
| description | Use when you hit any bug, test failure, or unexpected behavior - find the root cause before proposing a fix. |
Guessing wastes time and adds bugs. Find the root cause first, then fix it once.
Core principle: No fix before you understand the cause. A symptom patch is a failure, even when it makes the red go away.
Any test failure, crash, wrong output, performance cliff, build break, or "that's weird" - especially under time pressure, when a quick fix looks obvious, or when a previous fix did not hold. Simple bugs have root causes too.
Investigate. Read the error and stack trace completely - they often name the cause. Reproduce it reliably; if you cannot, gather more data rather than guess. Check what changed recently (git diff, recent commits, new deps/config). In a multi-component path, add instrumentation at each boundary and run once to see where it breaks before deciding why.
Trace to the source. Follow the bad value backward - where did it originate, what passed it in - until you reach the origin. Fix at the source, not where the symptom surfaced. Compare against a working example in the same codebase and list every difference; do not assume "that cannot matter".
One hypothesis at a time. State it: "I think X is the cause because Y." Make the smallest change that tests it - one variable. Did it work? If not, form a new hypothesis; do not stack fixes. If you do not understand something, say so and dig, do not pretend.
Fix at the root, with a test. Write a failing test that reproduces the bug and watch it fail (the test-first discipline from implementing). Make one root-cause fix, then verify: the test passes, nothing else broke, the issue is actually gone.
After 3 failed fixes, stop fixing. The pattern - each fix uncovers new coupling elsewhere, or needs "massive refactoring" - means the architecture is wrong, not the hypothesis. Raise it with your human partner instead of attempting fix #4.
"Quick fix now, investigate later" · "just try changing X" · several changes at once · "skip the test, I'll check by hand" · "it's probably X" · listing fixes before tracing the data flow · "one more attempt" after two failures.
