本文へ移動
cccskills
無料GitHub で公開

bug-report-writing

Use when any case fails or you reject a criterion - what goes in the review_criterion note, the failed case and the numbered need_revision comment, with severity

インストール方法を見る

含まれるファイル(1)

  • SKILL.md3.2 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Bug Report Writing

A defect's evidence lives in three places, each with its own job — never duplicate the full report into all three:

PlaceContents
Failed caseactual = the observed output, verbatim. evidence = exact command/flow plus frequency, e.g. "3/3".
Rejected criterion noteExpected (quote the criterion), actual, reproduction, environment: short SHA, URL/base, width or device, browser/runtime.
need_revision commentA numbered index, one line per defect: 1. [AC2 · case "export rejects >10k rows"] 500 instead of 422 — S2. Keep it inside the 15-line comment budget (concise-board-comments) — the reproduction lives in the criterion note and case evidence, not here.

What fails the task

A task fails when:

  • (a) a criterion or a human requirement comment is unmet;
  • (b) the team UI floor is broken, by a measured threshold: horizontal overflow at any of 360/768/1024/1440, a touch target <44px at 360, an input font <16px at 360, body contrast <4.5:1, a missing loading/empty/error state on a changed view, an uncaught console error or failed request in the flow, a broken image/icon, a critical/serious axe violation inside a changed screen;
  • (c) a crash, data loss, a security or authz hole, or a regression of adjacent behaviour;
  • (d) a Blocker or High deviation from the approved design the task builds (design-conformance) — Medium and Nitpick are notes.

Anything else — taste, polish beyond the floor — is a note on the criterion, never a failure.

Rules

  • Title names the problem, not the fix, under 60 characters: "Export of >10k rows returns 500", not "Fix export pagination".
  • Isolate to the first failing step; give the minimal reproduction, not the whole scenario.
  • Separate observation from speculation. A location from a stack trace or a grep_code hit is fine (the named debugging exception); a proposed fix is not your job.
  • Severity, worst wins: S1 data loss/security hole > S2 broken flow (criterion unmet) > S3 degraded UX (UI-floor violation) > S4 cosmetic (never fails the task on its own — see qa-criterion-verdicts).
  • Never invent a file path. A screenshot has none; describe width/URL/what-it-showed instead (ui-visual-evidence).
  • One defect per case/criterion entry — do not bundle unrelated failures into one note.

Worked Example

❌ "Export broken" — no repro, no expected, no severity.

✅

Case: export-pagination-10k (failed, 3/3)
actual: HTTP 500, body {"error":"internal"}
evidence: curl -X GET "$BASE/api/export?limit=10001" -H "Authorization: Bearer $TOKEN" — ran 3x, same result

Criterion note (AC2, rejected):
Expected: "requests over the 10,000-row limit return 422 with a clear message" (AC2)
Actual: 500 with no message, see case export-pagination-10k
Repro: curl -X GET "$BASE/api/export?limit=10001" ...
Env: SHA a1b2c3d, http://localhost:5173, desktop 1440

need_revision comment:
1. [AC2 · case "export-pagination-10k"] 500 instead of 422 at limit+1 — S2

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Use when writing acceptance criteria for a task - express each as an observable Given/When/Then that QA can execute, including negative cases

日本語の概要は準備中です。原文の説明を表示しています。

makifbaysal/tasktrooper1122026年10月11日 更新

Use when the diff adds or changes an endpoint, resolver, RPC, job or query that takes an object id, a role check, a request binding or a tenant filter - BOLA/IDOR, function-level authorization, mass assignment and tenant scoping

日本語の概要は準備中です。原文の説明を表示しています。

makifbaysal/tasktrooper1122026年10月11日 更新

Use on every UI change - semantic HTML, labels for controls, keyboard-navigable dialogs/menus, visible focus, and never color as the only signal

日本語の概要は準備中です。原文の説明を表示しています。

makifbaysal/tasktrooper1122026年10月11日 更新

Use when a task changes any screen, form, dialog, menu or control - Lighthouse/axe scan of the changed screens, a keyboard walk, and the thresholds that fail a task

日本語の概要は準備中です。原文の説明を表示しています。

makifbaysal/tasktrooper1122026年10月11日 更新

How to work a task returned with review, QA or UAT findings. Use when a task is in need_revision or PR review comments are in your context.

日本語の概要は準備中です。原文の説明を表示しています。

makifbaysal/tasktrooper1122026年10月11日 更新

Use when deciding whether a request needs an analiz task before implementation - the conditions that require the architect's analysis versus going straight to implementation

日本語の概要は準備中です。原文の説明を表示しています。

makifbaysal/tasktrooper1122026年10月11日 更新

makifbaysal のスキルをすべて見る

このスキルの問題を報告する