[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M72THXAPV8HCYCBN9XHXT64X":3,"skill-same:01M72THXAPV8HCYCBN9XHXT64X":83},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M72THXAPV8HCYCBN9XHXT64X","ywc-gen-testcase",null,"(ywc) Use when generating a manual verification testsheet (developer + QA gate) from a PR, task directory, task directory range, git range, or diff, with Markdown or HTML output. Triggers: \"generate test case\", \"testsheet\", \"QA checklist\", \"PR 테스트 케이스\", \"테스트시트\", \"テストシート作って\", \"manual test\", \"release 범위 test\", \"수기 검증\". Do not use for writing automated unit\u002Fintegration tests, code-level test generation, or in-IDE test scaffolding.","yongwoon\u002Fywc-agent-toolkit","yongwoon","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F9317029?v=4","01M72SZSF6VKANKG5QTMV4V95S","01M72SZSG8HHC3W2CV5P7EJ5AH",8,"MIT","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FSKILL.md",[],[],"2026-11-10 22:39:42",false,"---\nname: ywc-gen-testcase\ndescription: >-\n  (ywc) Use when generating a manual verification testsheet (developer + QA gate) from a PR, task directory, task directory range, git range, or diff, with Markdown or HTML output. Triggers: \"generate test case\", \"testsheet\", \"QA checklist\", \"PR 테스트 케이스\", \"테스트시트\", \"テストシート作って\", \"manual test\", \"release 범위 test\", \"수기 검증\". Do not use for writing automated unit\u002Fintegration tests, code-level test generation, or in-IDE test scaffolding.\n---\n\n# Generate Testsheet\n\n**Announce at start:** \"I'm using the ywc-gen-testcase skill to produce a dual-audience (developer + QA) testsheet.\"\n\nProduce a **dual-audience** testsheet (Markdown by default, or interactive HTML with `--format html`) from a GitHub PR, an implemented task directory, a task directory range, a git range, or the current git diff. The testsheet cleanly separates what the **developer** does (DB queries, curl contracts, container logs) from what a **QA\u002FPM\u002FProduct Owner** does in a browser, so both gates — pre-merge and pre-release — can be signed off independently and in parallel.\n\n## Rationalization Defense\n\nWhen tempted to skip a step, check this table first:\n\n| Excuse | Reality |\n|---|---|\n| \"Two input modes given, pick the more specific one silently\" | Stop and ask. Inputs cover different scopes (a PR vs a diff). Wrong choice produces a testsheet with mismatched scope. |\n| \"Developer test cases are obvious, focus on QA cases\" | Both audiences need explicit cases. Pre-merge gate (dev) and pre-release gate (QA) must be parallel-signable. |\n| \"Mix dev curl checks into the QA section for completeness\" | Each audience must own its section. Cross-contamination breaks parallel sign-off. |\n| \"Testsheet output without prerequisites\u002Fsetup is fine\" | Prerequisites (env, credentials, DB state) must be explicit. A testsheet that cannot be started is not a testsheet. |\n| \"Negative test cases (failure paths) are out of scope\" | Always include critical negative cases. Happy-path-only testsheets miss most regressions. |\n| \"Browser-dependent steps belong in dev section\" | UI verification belongs to QA. Dev section stays in terminal\u002Fcurl\u002FDB. |\n| \"Expected: 'the feature works correctly' is clear enough\" | Weak criteria require constant clarification. Expected must name concrete values: status codes, row counts, UI text, exact error messages. |\n\n**Violating the letter of these rules is violating the spirit.** A testsheet that mixes audiences is sign-off theater.\n\n## Inputs\n\nExactly one input mode must resolve. If more than one is given, stop and ask — they cover different scopes (a PR spans many commits; a diff spans only the current working tree), and silently picking one would produce a testsheet whose title, scope, and sign-off surface do not match the user's intent.\n\n| Input mode | How to specify | Primary source |\n|---|---|---|\n| **PR** | PR URL or PR number | `gh pr view`, `gh pr diff` |\n| **Task** | Task name or `\u003Cphase>-\u003Csequence>` prefix | `\u003Ctasks-dir>\u002F\u003Ctask>\u002Ftask.md` + `README.md` |\n| **Task Range** | Positional `\u003Ctask-start>..\u003Ctask-end>` | Inclusive ordered task directories from `\u003Ctasks-dir>` |\n| **Range** | Positional `\u003Cstart>..\u003Cend>` (SHA \u002F tag \u002F branch \u002F `HEAD~N`) or `--range \u003Cspec>` | `git log`, `git diff \u003Cstart>..\u003Cend>` |\n| **Diff** | `--from-diff` flag | Current `git diff HEAD` |\n\nIf no input is resolvable, stop and ask. Do not guess.\n\n## Arguments\n\nParse `$ARGUMENTS`.\n\n| Parameter | Format | Example | Description |\n|---|---|---|---|\n| PR identifier | URL or number | `250` | Fetches PR metadata + diff via `gh` |\n| Task specifier | name or prefix | `000001-010` | Prefix match against `\u003Ctasks-dir>` |\n| Task range | `\u003Ctask-start>..\u003Ctask-end>` | `000012-010..000019-010` | Inclusive range of ordered task directories. Resolve before treating positional `A..B` as a git range |\n| `--from-diff` | flag | | Use `git diff HEAD` |\n| `--range \u003Cspec>` | `\u003Cstart>..\u003Cend>` | `--range v1.2..v1.3` | Explicit git range form. Two-dot only — three-dot `A...B` is rejected |\n| `--output-dir \u003Cpath>` | path | `--output-dir .\u002Fdocs\u002Fqa` | Override `docs\u002Ftest-case\u002F` |\n| `--lang \u003Ccode>` | `ja`,`ko`,`en`,`zh`,`es` | `--lang zh` | Prose language. If omitted, resolve through shared YWC language policy. |\n| `--filename \u003Cname>` | string | `--filename release-smoke` | Override derived filename (no `.md`) |\n| `--tasks-dir \u003Cpath>` | path | `--tasks-dir .\u002Fdocs\u002Ftasks` | Tasks directory (Task and Task Range input only) |\n| `--include-regression` | flag | | Add Regression section (B side) |\n| `--audience \u003Cwho>` | `dev`\\|`qa`\\|`both` | `--audience qa` | Generate only one audience. Default - `both` (single file, A+B sections) |\n| `--split` | flag | | Force physical split into `\u003Cslug>-dev.md` + `\u003Cslug>-qa.md`. Mutually exclusive with `--force-single` |\n| `--force-single` | flag | | Bypass the L-tier split suggestion and always produce one file |\n| `--no-toc` | flag | | Suppress TOC auto-insertion for M\u002FL tier |\n| `--dry-run` | flag | | Show plan (tier, filenames, section counts) without writing |\n| `--format` | `markdown`\\|`html` | `--format html` | Output format. Default `markdown`. With `html`, writes an interactive HTML testsheet (see Step 5). |\n\n**Flag conflicts**: `--split` and `--force-single` cannot coexist — stop and ask. `--audience dev|qa` implies a single-audience file; `--split` in that combination is redundant and is silently ignored. **Git range input** is mutually exclusive with PR, Task, Task Range, and `--from-diff` — stop and ask if more than one is given. Range accepts only the two-dot `A..B` form; three-dot `A...B` is rejected because its merge-base semantics silently change scope and surprise the tester.\n\n## Context\n\n- Repository: !`gh repo view --json nameWithOwner --jq .nameWithOwner 2>\u002Fdev\u002Fnull || echo \"not a gh repo\"`\n- Current branch: !`git branch --show-current`\n- Changed files (diff mode reference): !`git status --short`\n- Default output directory: `docs\u002Ftest-case\u002F`\n\n## Pre-flight\n\n1. **Inside a repository** — `git rev-parse --show-toplevel` must succeed; run subsequent commands from the repo root.\n2. **`gh` CLI authenticated** (PR input only) — `gh auth status` must succeed.\n3. **Output directory writable** — `mkdir -p \u003Coutput-dir>` must succeed.\n4. **Exactly one input mode resolved** — if more than one input is present, stop and ask.\n5. **Flag conflicts** — see Arguments note above.\n\n## Step 1: Input Resolution\n\nResolve input mode in this order:\n\n1. Explicit PR, task, `--from-diff`, and `--range` flags\u002Fphrases.\n2. For positional two-dot `A..B`, reject `A...B` first.\n3. For positional two-dot `A..B`, check whether either endpoint is task-like (for example, starts with `000012-010`) or resolves as a task prefix under `\u003Ctasks-dir>`. If yes, use **Task Range** mode. Do not run `git rev-parse` before this task-range check.\n4. If the positional two-dot endpoints are not task-like and do not resolve as task prefixes, use **Git Range** mode.\n\n`--range A..B` always means Git Range. Use it when a branch, tag, or SHA happens to look like a task prefix.\n\n### 1a. PR input\n\n```bash\ngh pr view \u003CN> --json number,title,body,author,headRefName,baseRefName,url,files,commits,labels\ngh pr diff \u003CN>\n```\n\nCapture PR URL, title, body, author, head\u002Fbase branches, labels, changed files, commit messages. Lift any explicit \"How to test\" \u002F \"Acceptance Criteria\" \u002F \"動作確認\" block from the PR body into scenario goals verbatim.\n\n### 1b. Task input\n\n- Determine `\u003Ctasks-dir>` from `--tasks-dir`; if omitted, use the first existing directory among `tasks\u002F`, `docs\u002Ftasks\u002F`, `docs\u002Fspecification\u002Ftasks\u002F`, and `docs\u002Fspecifications\u002Ftasks\u002F`.\n- Resolve the task directory via prefix match. Prefer `\u003Ctasks-dir>\u002Fcompleted\u002F\u003Cname>\u002F` over `\u003Ctasks-dir>\u002F\u003Cname>\u002F` — a testsheet after merge is the common case.\n- Read `task.md` (Implementation Steps, Acceptance Criteria, Verify commands) and `README.md` in full.\n\n### 1c. Task Range input\n\nUse this mode for positional task ranges such as `000012-010..000019-010`.\n\n- Resolve each endpoint with the same prefix-match rule as Task input.\n- Build the ordered task list from directory basenames under `\u003Ctasks-dir>`, including common status buckets such as `completed\u002F`. Sort by basename lexicographically; numbered task prefixes are designed to sort in execution order.\n- Include every task from the resolved start task through the resolved end task, inclusive.\n- If either endpoint is missing or ambiguous, report the missing\u002Fambiguous task endpoint and stop. Do not fall back to Git Range for task-like endpoints.\n- If the start task appears after the end task in the ordered list, stop and ask whether the user intended the reverse range. Do not silently swap endpoints.\n- Read each selected task's `task.md` and `README.md` in full. Use task Acceptance Criteria and Verify commands as the primary scenario source, grouped by task in the Summary and Appendix when needed.\n\n### 1d. Diff input\n\n- `git diff HEAD` plus `git log \u003Cbase>..HEAD --oneline` (auto-detect `\u003Cbase>`: `develop` > `main` > `master`). Stop on empty diff.\n\n### 1e. Git Range input\n\nPositional `\u003Cstart>..\u003Cend>` or `--range \u003Cspec>`. Each endpoint accepts SHA, tag, branch, or a relative ref like `HEAD~5`. Reject the three-dot form `A...B` explicitly.\n\nCollect:\n\n- `git rev-parse \u003Cstart>` and `git rev-parse \u003Cend>` — fail fast on unknown refs\n- `git log --oneline \u003Cstart>..\u003Cend>` — commit list (used as scenario source)\n- `git diff \u003Cstart>..\u003Cend>` — the change set (used for surface classification)\n- Stop immediately on an empty range\n\n### 1f. Range quality checks (git range input only)\n\nRange mode lacks PR body \u002F Task Acceptance Criteria, so scenario quality is sensitive to commit-message hygiene and scope size. Three protective checks run before Step 2 to prevent low-quality testsheets from being generated silently:\n\n1. **PR auto-suggest** — run `gh pr list --search \u003Chead-sha> --state all --json number,title,commits` on the range's HEAD commit. If an open or merged PR exists and its commit set covers ≥ 80% of the range, stop and ask:\n\n   ```text\n   Range HEAD is part of PR #\u003CNNN> \"\u003Ctitle>\" (covers N of M commits).\n   PR mode provides PR body, labels, and explicit \"How to test\" context —\n   scenario quality is meaningfully better.\n\n       [a] Switch to PR #\u003CNNN> mode (recommended)\n       [b] Keep range mode (PR metadata will be ignored)\n   ```\n\n2. **Commit-message quality** — if ≥ 70% of commit headlines in the range are ≤ 10 characters (e.g. `wip`, `fix`, `tmp`, `update`), warn and ask whether to proceed. Low-signal headlines force scenario generation to rely on diff inference alone, which is materially weaker than structured acceptance criteria.\n\n3. **Range size guard** — if the range contains > 50 commits, escalate directly to L-tier handling in Step 4 regardless of scenario count. Review burden grows with commit lineage length, not just with scenario count.\n\n## Step 2: Audience and Surface Classification\n\nEvery scenario carries **two** tags: **audience** (who runs it) and **surface** (what it verifies). Classifying on both axes is what makes the final testsheet useful to different readers.\n\n### 2a. Audience tagging\n\n| Audience | Who | Tools | Gate |\n|---|---|---|---|\n| **A. Developer** | Backend \u002F DB \u002F DevOps engineer | `psql`, `gh`, `curl`, `docker`, `kubectl`, unit\u002Fintegration test runners | **Pre-merge** — contracts, migrations, worker runners, container orchestration |\n| **B. QA \u002F Browser** | QA, PM, Product Owner, designer | Chrome + DevTools, admin UI, test HTML origin, mobile device\u002Femulator | **Pre-release** — end-user experience and any browser-observable effect |\n\nScenarios that require both (e.g. \"SDK loader works end-to-end\") belong to **B**, with the data setup moved into A's Prerequisites. This prevents duplicate work.\n\n### 2b. Surface classification\n\n| Surface | File patterns (typical) | Scenarios produced | Typical audience |\n|---|---|---|---|\n| UI \u002F Browser | `*.tsx`,`*.vue`,`*.svelte`,`app\u002F**`,`pages\u002F**`,`components\u002F**`,`*.css`,`*.html` | Route\u002Fscreen walkthrough, form validation, empty\u002Floading\u002Ferror states | B |\n| Database \u002F Table | `migrations\u002F**`,`*.sql`,`prisma\u002Fschema.prisma`,`supabase\u002Fmigrations\u002F**` | Row visibility, constraint enforcement, index presence, rollback safety | A |\n| API | `app\u002Fapi\u002F**`,`*.controller.*`,`**\u002Froute.ts`,`src\u002Froutes\u002F**`,`api\u002F**\u002F*.py` | Status codes, response shape, auth boundary, validation errors | A primarily; B when endpoint is browser-consumed |\n| Background job \u002F Webhook | `workers\u002F**`,`jobs\u002F**`,`webhook*`,`cron*`,`*.worker.ts` | Trigger condition, retry, idempotency | A |\n| Configuration | `*.env*`,`*.yaml`,`*.toml`,`Dockerfile`,`docker-compose*`,`.github\u002Fworkflows\u002F**` | Env presence, build\u002Fruntime, CI green | A |\n| Docs only | `*.md`,`docs\u002F**` | Docs review checklist | A or B depending on audience of the doc |\n\n**Do not fabricate surfaces that do not exist in the diff.** An empty section tells the tester nothing and dilutes the real work.\n\n### 2c. Browser-observable effects (even in backend-only PRs)\n\nFile-pattern matching alone is not sufficient to decide the B section. A backend-only PR can still have browser-observable effects that only a real browser can verify — curl cannot reliably reproduce CORS preflight, cookie partitioning, `\u003Cscript>` tag loading, `Content-Type` sniffing, or downstream dashboard behaviour.\n\n**Include a B (QA\u002FBrowser) section even when no frontend files changed, if any of the following are true:**\n\n| Browser-observable effect | B-section scenario to include |\n|---|---|\n| Public endpoint consumed by a browser (`\u002Fconfig\u002F:id`, SDK loader, webhook receiver) | \"Fetch from DevTools Console\" scenario |\n| CORS \u002F preflight behaviour changed | Preflight-from-real-browser scenario (curl does not fully simulate preflight) |\n| `Content-Type` \u002F CSP \u002F cookie attributes changed | DevTools Network tab verification |\n| Backend APIs consumed by an **existing unchanged** frontend in the same repo | Consumer-renders-successfully scenario (regression-style) |\n| Script\u002Fasset delivered to end-user browsers (SDK, embed, widget) | \"Load from test HTML page\" scenario |\n\nThe test for whether to include a B section is not \"did a `.tsx` file change\" but **\"can a human verify this change by opening a browser?\"** When in doubt, include a minimal B section of 1–2 scenarios rather than omit it — a missing B section on a release that touches public endpoints is a common source of production surprises.\n\n## Step 3: Scenario Generation\n\nFor each (audience, surface) pair that exists, generate 2–5 scenarios. A scenario is one testable behaviour, not one changed file.\n\nEach scenario has:\n\n- **Goal** — one sentence\n- **Preconditions** — state the app must be in\n- **Steps** — numbered, imperative, specific enough for a tester who did not write the code\n- **Expected** — concrete values (`expect 3 rows with status = 'active'`), not \"it should work\"\n- **Checkbox** — one `- [ ]` at the top\n\n**Anti-pattern**: `- [ ] Test that the feature works`. That is not a scenario. If you cannot state *what to run\u002Fclick and what to see*, the scenario is not ready.\n\nScenario sources, in priority order:\n\n1. PR body \u002F task Acceptance Criteria (\"How to test\", \"動作確認\", \"Reviewer should verify\")\n2. Commit messages\n3. Surface-driven scenarios (from 2b)\n4. Diff edge cases (new `NOT NULL` → null-input scenario; new permission check → unauthorised-role scenario)\n\n### Triggered security negative paths\n\nRead the shared security omission lenses at `ywc-spec-validate\u002Freferences\u002Fsecurity-qa-lenses.md`, resolved from the skills root containing this installed skill (not the working directory), against the already-acquired diff, task\u002Fspec and verified project context. If the installed reference is missing or unreadable, return `NEEDS_CONTEXT` identifying that path and request installation of `ywc-spec-validate`; do not silently skip security checks or invent a replacement rubric. Add one negative-path scenario for each triggered lens, with concrete preconditions, steps and observable denial\u002Fsafe-failure outcome. Deduplicate overlapping scenarios without dropping any triggered check: record all covered lens names on the retained scenario and preserve each distinct expected outcome. These security scenarios are in addition to the 2–5 per (audience, surface) pair above, so that limit never forces a triggered lens to be dropped; they still count toward the Step 4 scenario count. Do not invent untriggered security surfaces; record N\u002FA reasons where appropriate. Scope manual checks to safe test fixtures and authorized environments; this does not authorize attacks or production probes.\n\nReuse evidence from the selected input mode. Do not fetch a PR body solely to import Impact; Impact already present in acquired evidence may inform scenarios but never substitutes for verified scope. Do not add unconditional PR-body acquisition, another diff read, screenshot operations or broad collection for this feature.\n\n## Step 4: Tier Decision\n\nCount scenarios after Step 3. Use the count to choose layout. This protects testsheets from becoming unread 1000-line walls.\n\n| Tier | Scenarios | Layout |\n|---|---|---|\n| **S** | ≤ 20 | Single file, A+B sections, **no TOC**, **no collapsible details** |\n| **M** | 21–40 | Single file, A+B sections, **auto-TOC** at top, **`\u003Cdetails>` collapsible** for Prerequisites + Edge Cases |\n| **L** | > 40 scenarios **OR** range > 50 commits | **Ask user** before writing:\u003Cbr>1. Single file with TOC + collapsible (as M)\u003Cbr>2. `--split` into `\u003Cslug>-dev.md` + `\u003Cslug>-qa.md`\u003Cbr>3. Split by Phase \u002F feature (user names each file)\u003Cbr>Proceed only after the user chooses. `--force-single` bypasses the prompt. |\n\nExplicit flags override the tier default -\n\n- `--split` → physical split, regardless of tier\n- `--audience dev|qa` → single-audience file, skip the other section entirely\n- `--force-single` → one file even if count > 40\n- `--no-toc` → suppress TOC even in M\u002FL\n\n## Step 5: Write the Testsheet\n\n### Output format\n\nDefault is a Markdown testsheet (`.md`) — keep this for the existing `-v\u003CN>`\nappend and front-matter sign-off workflow. With `--format html`, write an\n**interactive HTML testsheet** instead, following\n[html-output.md](..\u002Freferences\u002Fhtml-output.md): each scenario checkbox is\nclickable, sign-off state (tester name, Pass\u002FFail\u002FBlocked, notes) persists in\nthe browser via `localStorage`, and a `Copy as Markdown` button exports the\ncurrent sign-off state back to the Markdown testsheet format so the result can\nbe pasted into a PR or committed. HTML mode is recommended when the testsheet\nwill be handed to a QA\u002FPM who signs off in a browser; the filename follows the\nDefault filename table below with an `.html` extension. The Markdown surface\n(front matter included) is preserved inside the file.\n\n### Default filename\n\n| Input | Single file (default) | `--split` mode |\n|---|---|---|\n| PR | `pr-\u003Cnumber>-\u003Cslug>.md` | `pr-\u003Cnumber>-\u003Cslug>-dev.md` + `...-qa.md` |\n| Task | `task-\u003Cphase>-\u003Csequence>-\u003Cslug>.md` | `...-dev.md` + `...-qa.md` |\n| Task Range | `tasks-\u003Cstart-prefix>-\u003Cend-prefix>-\u003Cslug>.md` | `...-dev.md` + `...-qa.md` |\n| Range | `range-\u003Cshort-start>-\u003Cshort-end>-\u003Cslug>.md` (tag names used verbatim when both endpoints are tags, e.g. `range-v1.2-v1.3-\u003Cslug>.md`) | `...-dev.md` + `...-qa.md` |\n| Diff | `\u003Cyyyymmdd-HHMM>-\u003Cbranch-slug>.md` | `...-dev.md` + `...-qa.md` |\n\nIf the target file already exists, append `-v\u003CN>` (never overwrite). Existing files may carry a tester's partial sign-off.\n\n### Single-file template (default, A+B)\n\nWrite `\u003Coutput-dir>\u002F\u003Cfilename>.md` using this skeleton. Keep front matter keys, section numbers, file names, code snippets, and template structure in English regardless of `--lang`; they are machine surfaces. M\u002FL tier inserts a TOC block and wraps Prerequisites\u002FEdge Cases in `\u003Cdetails>`; S tier skips both.\n\n````markdown\n---\nsource: \u003Cpr#N | task:\u003Cname> | task-range:\u003Cstart>..\u003Cend> | diff | range>\npr_url: \u003Curl or \"\">\ntask_range_spec: \u003Ce.g. \"000012-010..000019-010\" or \"\">\ntask_count: \u003Cnumber or \"\">\nrange_spec: \u003Ce.g. \"main..feature-x\" or \"\">\nrange_start: \u003Cshort SHA or tag, or \"\">\nrange_end: \u003Cshort SHA or tag, or \"\">\nbranch: \u003Chead branch, or \"\" for ranges crossing branches>\nbase_branch: \u003Cbase>\nauthor: \u003Cusername>\ngenerated_at: \u003CISO 8601>\ntier: \u003CS | M | L>\ndev_tester: \"\"\ndev_status: pending\nqa_tester: \"\"\nqa_status: pending\n---\n\n# Test Case: \u003CTitle>\n\n\u003C!-- TOC block: M\u002FL tier only -->\n## Table of Contents\n- [1. Summary](#1-summary)\n- [2. Prerequisites](#2-prerequisites)\n- [A. Developer Verification](#a-developer-verification)\n  - [A.1 Database](#a1-database--table)\n  - [A.2 API](#a2-api)\n  - [A.3 Background Jobs](#a3-background-jobs--workers)\n  - [A.4 Configuration](#a4-configuration)\n  - [A.5 Dev Edge Cases](#a5-dev-edge-cases)\n  - [A.6 Dev Sign-off](#a6-dev-sign-off)\n- [B. QA \u002F Browser Verification](#b-qa--browser-verification)\n  - [B.1 UI \u002F Browser Scenarios](#b1-ui--browser-scenarios)\n  - [B.2 User-visible Edge Cases](#b2-user-visible-edge-cases)\n  - [B.3 Regression](#b3-regression)\n  - [B.4 QA Sign-off](#b4-qa-sign-off)\n\n## 1. Summary\n\n\u003C2–4 sentences explaining what changed and why a human needs to verify it.>\n\n## 2. Prerequisites\n\n\u003C!-- M\u002FL tier: wrap in \u003Cdetails> tag, open by default -->\n\n### 2.0 Common (both audiences)\n- [ ] …\n\n### 2.A Dev-only\n- [ ] …\n\n### 2.B QA-only\n- [ ] …\n\n## A. Developer Verification\n\n> **Audience**: Backend \u002F DB \u002F DevOps engineer  \n> **Tools**: psql, gh, curl, docker  \n> **Gate**: pre-merge\n\n### A.1 Database \u002F Table\n\u003Conly if surface exists>\n\n#### A.1.1 \u003Cscenario name>\n- [ ] **Goal**: …\n- **Preconditions**: …\n- **Steps**:\n  1. …\n- **Expected**: …\n\n### A.2 API\n### A.3 Background Jobs \u002F Workers\n### A.4 Configuration\n### A.5 Dev Edge Cases\n\n\u003C!-- M\u002FL tier: wrap A.5 in \u003Cdetails> -->\n\n- [ ] \u003Cboundary \u002F null \u002F error path>\n\n### A.6 Dev Sign-off\n- **Tester**:\n- **Date**:\n- **Result**: ☐ Pass ☐ Fail ☐ Blocked\n- **Notes**:\n\n## B. QA \u002F Browser Verification\n\n> **Audience**: QA \u002F PM \u002F Product Owner \u002F Designer  \n> **Tools**: Chrome + DevTools, admin UI, test origin  \n> **Gate**: pre-release\n\n### B.1 UI \u002F Browser Scenarios\n\u003Conly if a browser-observable surface exists>\n\n### B.2 User-visible Edge Cases\n\n\u003C!-- M\u002FL tier: wrap in \u003Cdetails> -->\n\n### B.3 Regression\n\u003Conly if --include-regression>\n\n### B.4 QA Sign-off\n- **Tester**:\n- **Date**:\n- **Result**: ☐ Pass ☐ Fail ☐ Blocked\n- **Notes**:\n\n## Appendix\n\u003Coptional — long SQL, payload samples, troubleshooting>\n````\n\n### Split-file variants (`--split` or `--audience`)\n\nEach file contains only that audience's Summary, Prerequisites, Scenarios, and Sign-off. Add a **Cross-reference** line at the bottom pointing to the counterpart (`Developer sheet: .\u002F\u003Cslug>-dev.md`, `QA sheet: .\u002F\u003Cslug>-qa.md`). Omit cross-reference when `--audience` was explicit and only one file is produced.\n\n## Length Management Guidelines\n\nApply these regardless of tier to keep testsheets lean:\n\n1. **Prerequisites common-prefix + audience-suffix** — shared rows live once in `## 2.0 Common`; dev-only \u002F qa-only rows live in `## 2.A` \u002F `## 2.B`. Never duplicate.\n2. **Extract long verification material** — any SQL \u002F script > 20 lines lives in `scripts\u002Fqa\u002F*.sql` (or similar) and the scenario references the path only. The scenario body stays short.\n3. **Regression by reference** — instead of repeating known-good regression flows, link to prior testsheets (e.g. `Stripe release testsheet: .\u002Fpr-230-stripe-release.md`).\n4. **Appendix at the bottom** — overflow content (detailed troubleshooting, sample payloads) goes under `## Appendix` and is linked from the relevant scenario.\n\nApplied consistently, these four rules keep most M-tier testsheets under ~800 lines and prevent L-tier by shifting volume out of the main flow.\n\n## Step 6: Validate and Report\n\nBefore finishing:\n\n1. File(s) exist and non-empty.\n2. Checkbox count ≥ 3.\n3. Every scenario has a concrete **Expected** result.\n4. If M\u002FL tier and `--no-toc` not set, TOC present at the top with anchors matching headings.\n5. If split layout, cross-reference line present in each file.\n\nReport:\n\n```text\nTestsheet generated: \u003Cabsolute path(s)>\nSource: \u003Cpr#N | task:\u003Cname> | task range:\u003Ctask-start>..\u003Ctask-end> | diff | git range:\u003Cstart>..\u003Cend>>\nTier: \u003CS | M | L>\nLayout: \u003Csingle | split>\nAudiences: \u003CA only | B only | A+B>\nScenarios by surface: UI (3), DB (1), API (2), Jobs (4), Config (2)\nScenarios by audience: Dev \u003Cn>, QA \u003Cm>\nTotal: \u003Cscenarios> scenarios, \u003Ccheckboxes> checkboxes\nLanguage: \u003Cdetected-or-specified>\nNext steps:\n  - Dev signs off on Section A (pre-merge gate)\n  - QA signs off on Section B (pre-release gate)\n  - Update front-matter `dev_status:` \u002F `qa_status:` as work progresses\n```\n\n## Language Detection\n\nWhen `--lang` is not specified, resolve testsheet prose language using the\nshared Codex YWC policy:\n[`..\u002Freferences\u002Flanguage-resolution.md`](..\u002Freferences\u002Flanguage-resolution.md).\nThe order is explicit `--lang` > project `.codex\u002Fywc.json` > project guidance\n(`AGENTS.md`, `CODEX.md`, `CLAUDE.md`) > user `~\u002F.codex\u002Fywc.json` > ask user.\n\nRecent testsheets in the output directory and the project `README.md` may be\nused only as non-authoritative context when asking the user. They do not create\na final fallback default.\n\nSupported language codes are `ja`, `ko`, `en`, `zh`, and `es`. Treat aliases such as `japanese`, `korean`, `english`, `chinese`, `Chinese (Simplified)`, `中文`, `spanish`, `espanol`, and `español` as requests for the matching code; `zh` \u002F `chinese` means Simplified Chinese unless the user explicitly asks for a different Chinese locale.\n\nYAML front matter keys, section numbers, file names, code snippets, command names, and the template skeleton stay in English regardless of `--lang`. Only prose (Summary, Goal, Steps, Expected, Notes, Edge Cases, and audience-facing test descriptions) follows the chosen language. Technical terms that are part of the testsheet contract or tooling surface, such as Developer Verification, QA \u002F Browser, PR, Task, Git Range, Expected, front matter, Status, and Sign-off, stay English in `zh` and `es` prose.\n\n## Dry Run Mode\n\nWith `--dry-run`, perform Steps 1–4 as usual, then print:\n\n- Resolved input mode and source\n- Tier decision (S\u002FM\u002FL) and scenario count\n- Target filename(s) and full output path(s)\n- Surface classification (which sections will exist)\n- Scenarios per (audience, surface) pair\n- Detected language\n\nDo **not** write any file. Exit after printing.\n\n## Error Handling\n\n| Situation | Behaviour |\n|---|---|\n| No input given | Stop, ask for PR \u002F task \u002F range \u002F `--from-diff` |\n| Multiple inputs given (PR + Range, Task + Range, Range + `--from-diff`, etc.) | Stop, ask which mode |\n| Three-dot range `A...B` passed | Reject and ask for the two-dot `A..B` form instead |\n| Task range endpoint missing or ambiguous | Report the endpoint and similar task names; stop without trying `git rev-parse` |\n| Task range start appears after end | Stop and ask whether the user intended the reverse range |\n| Range ref invalid (`git rev-parse` fails on `\u003Cstart>` or `\u003Cend>`) | Report which ref failed; stop |\n| Range empty (`\u003Cend>` has no commits absent from `\u003Cstart>`) | Report \"range is empty\"; stop |\n| Range > 50 commits | Escalate to L-tier handling (ask user how to proceed) |\n| Range HEAD belongs to an open\u002Fmerged PR (≥ 80% overlap) | Stop and ask whether to switch to PR mode |\n| Range has ≥ 70% low-quality commit headlines | Warn and ask whether to proceed |\n| `--split` and `--force-single` both set | Stop, ask which is intended |\n| `gh` not authenticated (PR input or range PR auto-suggest) | Ask user to run `gh auth login`; stop |\n| PR not found | Report the PR number and stop |\n| Task not found | List similar tasks (fuzzy match) and stop |\n| Empty diff (diff input) | Report \"nothing to test\" and stop |\n| Output directory not writable | Report failing path and stop (no silent fallback) |\n| Target file already exists | Append `-v\u003CN>` suffix; report new filename |\n| L tier detected and no explicit `--split`\u002F`--force-single` | Ask user how to proceed before writing |\n\n## Integration\n\nDownstream of the implementation-oriented `ywc` skills:\n\n- **ywc-sequential-executor** — generates the PR\u002FTask this skill tests\n- **ywc-parallel-executor** — same, for parallel execution\n- **ywc-merge-dependabot** — merged dependency bumps that often warrant a smoke testsheet\n\nInvoke this skill after those finish, when a human tester is ready to validate.\n\n## Examples\n\nFor invocation examples across PR, task, task range, diff, and git range modes, read [references\u002Fexamples.md](references\u002Fexamples.md).\n\n## Output Format\n\nInclude `Status: \u003CDONE | DONE_WITH_CONCERNS | BLOCKED | NEEDS_CONTEXT>`, the generated testsheet path or dry-run summary, detected scope, and any blocker.\n\n## Validation\n\nBefore finalizing, verify that input mode is unambiguous, the diff\u002Frange is non-empty, output path handling is deterministic, and no file is written during `--dry-run`.\n","# Generate Testsheet\n\n**Announce at start:** \"I'm using the ywc-gen-testcase skill to produce a dual-audience (developer + QA) testsheet.\"\n\nProduce a **dual-audience** testsheet (Markdown by default, or interactive HTML with `--format html`) from a GitHub PR, an implemented task directory, a task directory range, a git range, or the current git diff. The testsheet cleanly separates what the **developer** does (DB queries, curl contracts, container logs) from what a **QA\u002FPM\u002FProduct Owner** does in a browser, so both gates — pre-merge and pre-release — can be signed off independently and in parallel.\n\n## Rationalization Defense\n\nWhen tempted to skip a step, check this table first:\n\n| Excuse | Reality |\n|---|---|\n| \"Two input modes given, pick the more specific one silently\" | Stop and ask. Inputs cover different scopes (a PR vs a diff). Wrong choice produces a testsheet with mismatched scope. |\n| \"Developer test cases are obvious, focus on QA cases\" | Both audiences need explicit cases. Pre-merge gate (dev) and pre-release gate (QA) must be parallel-signable. |\n| \"Mix dev curl checks into the QA section for completeness\" | Each audience must own its section. Cross-contamination breaks parallel sign-off. |\n| \"Testsheet output without prerequisites\u002Fsetup is fine\" | Prerequisites (env, credentials, DB state) must be explicit. A testsheet that cannot be started is not a testsheet. |\n| \"Negative test cases (failure paths) are out of scope\" | Always include critical negative cases. Happy-path-only testsheets miss most regressions. |\n| \"Browser-dependent steps belong in dev section\" | UI verification belongs to QA. Dev section stays in terminal\u002Fcurl\u002FDB. |\n| \"Expected: 'the feature works correctly' is clear enough\" | Weak criteria require constant clarification. Expected must name concrete values: status codes, row counts, UI text, exact error messages. |\n\n**Violating the letter of these rules is violating the spirit.** A testsheet that mixes audiences is sign-off theater.\n\n## Inputs\n\nExactly one input mode must resolve. If more than one is given, stop and ask — they cover different scopes (a PR spans many commits; a diff spans only the current working tree), and silently picking one would produce a testsheet whose title, scope, and sign-off surface do not match the user's intent.\n\n| Input mode | How to specify | Primary source |\n|---|---|---|\n| **PR** | PR URL or PR number | `gh pr view`, `gh pr diff` |\n| **Task** | Task name or `\u003Cphase>-\u003Csequence>` prefix | `\u003Ctasks-dir>\u002F\u003Ctask>\u002Ftask.md` + `README.md` |\n| **Task Range** | Positional `\u003Ctask-start>..\u003Ctask-end>` | Inclusive ordered task directories from `\u003Ctasks-dir>` |\n| **Range** | Positional `\u003Cstart>..\u003Cend>` (SHA \u002F tag \u002F branch \u002F `HEAD~N`) or `--range \u003Cspec>` | `git log`, `git diff \u003Cstart>..\u003Cend>` |\n| **Diff** | `--from-diff` flag | Current `git diff HEAD` |\n\nIf no input is resolvable, stop and ask. Do not guess.\n\n## Arguments\n\nParse `$ARGUMENTS`.\n\n| Parameter | Format | Example | Description |\n|---|---|---|---|\n| PR identifier | URL or number | `250` | Fetches PR metadata + diff via `gh` |\n| Task specifier | name or prefix | `000001-010` | Prefix match against `\u003Ctasks-dir>` |\n| Task range | `\u003Ctask-start>..\u003Ctask-end>` | `000012-010..000019-010` | Inclusive range of ordered task directories. Resolve before treating positional `A..B` as a git range |\n| `--from-diff` | flag | | Use `git diff HEAD` |\n| `--range \u003Cspec>` | `\u003Cstart>..\u003Cend>` | `--range v1.2..v1.3` | Explicit git range form. Two-dot only — three-dot `A...B` is rejected |\n| `--output-dir \u003Cpath>` | path | `--output-dir .\u002Fdocs\u002Fqa` | Override `docs\u002Ftest-case\u002F` |\n| `--lang \u003Ccode>` | `ja`,`ko`,`en`,`zh`,`es` | `--lang zh` | Prose language. If omitted, resolve through shared YWC language policy. |\n| `--filename \u003Cname>` | string | `--filename release-smoke` | Override derived filename (no `.md`) |\n| `--tasks-dir \u003Cpath>` | path | `--tasks-dir .\u002Fdocs\u002Ftasks` | Tasks directory (Task and Task Range input only) |\n| `--include-regression` | flag | | Add Regression section (B side) |\n| `--audience \u003Cwho>` | `dev`\\|`qa`\\|`both` | `--audience qa` | Generate only one audience. Default - `both` (single file, A+B sections) |\n| `--split` | flag | | Force physical split into `\u003Cslug>-dev.md` + `\u003Cslug>-qa.md`. Mutually exclusive with `--force-single` |\n| `--force-single` | flag | | Bypass the L-tier split suggestion and always produce one file |\n| `--no-toc` | flag | | Suppress TOC auto-insertion for M\u002FL tier |\n| `--dry-run` | flag | | Show plan (tier, filenames, section counts) without writing |\n| `--format` | `markdown`\\|`html` | `--format html` | Output format. Default `markdown`. With `html`, writes an interactive HTML testsheet (see Step 5). |\n\n**Flag conflicts**: `--split` and `--force-single` cannot coexist — stop and ask. `--audience dev|qa` implies a single-audience file; `--split` in that combination is redundant and is silently ignored. **Git range input** is mutually exclusive with PR, Task, Task Range, and `--from-diff` — stop and ask if more than one is given. Range accepts only the two-dot `A..B` form; three-dot `A...B` is rejected because its merge-base semantics silently change scope and surprise the tester.\n\n## Context\n\n- Repository: !`gh repo view --json nameWithOwner --jq .nameWithOwner 2>\u002Fdev\u002Fnull || echo \"not a gh repo\"`\n- Current branch: !`git branch --show-current`\n- Changed files (diff mode reference): !`git status --short`\n- Default output directory: `docs\u002Ftest-case\u002F`\n\n## Pre-flight\n\n1. **Inside a repository** — `git rev-parse --show-toplevel` must succeed; run subsequent commands from the repo root.\n2. **`gh` CLI authenticated** (PR input only) — `gh auth status` must succeed.\n3. **Output directory writable** — `mkdir -p \u003Coutput-dir>` must succeed.\n4. **Exactly one input mode resolved** — if more than one input is present, stop and ask.\n5. **Flag conflicts** — see Arguments note above.\n\n## Step 1: Input Resolution\n\nResolve input mode in this order:\n\n1. Explicit PR, task, `--from-diff`, and `--range` flags\u002Fphrases.\n2. For positional two-dot `A..B`, reject `A...B` first.\n3. For positional two-dot `A..B`, check whether either endpoint is task-like (for example, starts with `000012-010`) or resolves as a task prefix under `\u003Ctasks-dir>`. If yes, use **Task Range** mode. Do not run `git rev-parse` before this task-range check.\n4. If the positional two-dot endpoints are not task-like and do not resolve as task prefixes, use **Git Range** mode.\n\n`--range A..B` always means Git Range. Use it when a branch, tag, or SHA happens to look like a task prefix.\n\n### 1a. PR input\n\n```bash\ngh pr view \u003CN> --json number,title,body,author,headRefName,baseRefName,url,files,commits,labels\ngh pr diff \u003CN>\n```\n\nCapture PR URL, title, body, author, head\u002Fbase branches, labels, changed files, commit messages. Lift any explicit \"How to test\" \u002F \"Acceptance Criteria\" \u002F \"動作確認\" block from the PR body into scenario goals verbatim.\n\n### 1b. Task input\n\n- Determine `\u003Ctasks-dir>` from `--tasks-dir`; if omitted, use the first existing directory among `tasks\u002F`, `docs\u002Ftasks\u002F`, `docs\u002Fspecification\u002Ftasks\u002F`, and `docs\u002Fspecifications\u002Ftasks\u002F`.\n- Resolve the task directory via prefix match. Prefer `\u003Ctasks-dir>\u002Fcompleted\u002F\u003Cname>\u002F` over `\u003Ctasks-dir>\u002F\u003Cname>\u002F` — a testsheet after merge is the common case.\n- Read `task.md` (Implementation Steps, Acceptance Criteria, Verify commands) and `README.md` in full.\n\n### 1c. Task Range input\n\nUse this mode for positional task ranges such as `000012-010..000019-010`.\n\n- Resolve each endpoint with the same prefix-match rule as Task input.\n- Build the ordered task list from directory basenames under `\u003Ctasks-dir>`, including common status buckets such as `completed\u002F`. Sort by basename lexicographically; numbered task prefixes are designed to sort in execution order.\n- Include every task from the resolved start task through the resolved end task, inclusive.\n- If either endpoint is missing or ambiguous, report the missing\u002Fambiguous task endpoint and stop. Do not fall back to Git Range for task-like endpoints.\n- If the start task appears after the end task in the ordered list, stop and ask whether the user intended the reverse range. Do not silently swap endpoints.\n- Read each selected task's `task.md` and `README.md` in full. Use task Acceptance Criteria and Verify commands as the primary scenario source, grouped by task in the Summary and Appendix when needed.\n\n### 1d. Diff input\n\n- `git diff HEAD` plus `git log \u003Cbase>..HEAD --oneline` (auto-detect `\u003Cbase>`: `develop` > `main` > `master`). Stop on empty diff.\n\n### 1e. Git Range input\n\nPositional `\u003Cstart>..\u003Cend>` or `--range \u003Cspec>`. Each endpoint accepts SHA, tag, branch, or a relative ref like `HEAD~5`. Reject the three-dot form `A...B` explicitly.\n\nCollect:\n\n- `git rev-parse \u003Cstart>` and `git rev-parse \u003Cend>` — fail fast on unknown refs\n- `git log --oneline \u003Cstart>..\u003Cend>` — commit list (used as scenario source)\n- `git diff \u003Cstart>..\u003Cend>` — the change set (used for surface classification)\n- Stop immediately on an empty range\n\n### 1f. Range quality checks (git range input only)\n\nRange mode lacks PR body \u002F Task Acceptance Criteria, so scenario quality is sensitive to commit-message hygiene and scope size. Three protective checks run before Step 2 to prevent low-quality testsheets from being generated silently:\n\n1. **PR auto-suggest** — run `gh pr list --search \u003Chead-sha> --state all --json number,title,commits` on the range's HEAD commit. If an open or merged PR exists and its commit set covers ≥ 80% of the range, stop and ask:\n\n   ```text\n   Range HEAD is part of PR #\u003CNNN> \"\u003Ctitle>\" (covers N of M commits).\n   PR mode provides PR body, labels, and explicit \"How to test\" context —\n   scenario quality is meaningfully better.\n\n       [a] Switch to PR #\u003CNNN> mode (recommended)\n       [b] Keep range mode (PR metadata will be ignored)\n   ```\n\n2. **Commit-message quality** — if ≥ 70% of commit headlines in the range are ≤ 10 characters (e.g. `wip`, `fix`, `tmp`, `update`), warn and ask whether to proceed. Low-signal headlines force scenario generation to rely on diff inference alone, which is materially weaker than structured acceptance criteria.\n\n3. **Range size guard** — if the range contains > 50 commits, escalate directly to L-tier handling in Step 4 regardless of scenario count. Review burden grows with commit lineage length, not just with scenario count.\n\n## Step 2: Audience and Surface Classification\n\nEvery scenario carries **two** tags: **audience** (who runs it) and **surface** (what it verifies). Classifying on both axes is what makes the final testsheet useful to different readers.\n\n### 2a. Audience tagging\n\n| Audience | Who | Tools | Gate |\n|---|---|---|---|\n| **A. Developer** | Backend \u002F DB \u002F DevOps engineer | `psql`, `gh`, `curl`, `docker`, `kubectl`, unit\u002Fintegration test runners | **Pre-merge** — contracts, migrations, worker runners, container orchestration |\n| **B. QA \u002F Browser** | QA, PM, Product Owner, designer | Chrome + DevTools, admin UI, test HTML origin, mobile device\u002Femulator | **Pre-release** — end-user experience and any browser-observable effect |\n\nScenarios that require both (e.g. \"SDK loader works end-to-end\") belong to **B**, with the data setup moved into A's Prerequisites. This prevents duplicate work.\n\n### 2b. Surface classification\n\n| Surface | File patterns (typical) | Scenarios produced | Typical audience |\n|---|---|---|---|\n| UI \u002F Browser | `*.tsx`,`*.vue`,`*.svelte`,`app\u002F**`,`pages\u002F**`,`components\u002F**`,`*.css`,`*.html` | Route\u002Fscreen walkthrough, form validation, empty\u002Floading\u002Ferror states | B |\n| Database \u002F Table | `migrations\u002F**`,`*.sql`,`prisma\u002Fschema.prisma`,`supabase\u002Fmigrations\u002F**` | Row visibility, constraint enforcement, index presence, rollback safety | A |\n| API | `app\u002Fapi\u002F**`,`*.controller.*`,`**\u002Froute.ts`,`src\u002Froutes\u002F**`,`api\u002F**\u002F*.py` | Status codes, response shape, auth boundary, validation errors | A primarily; B when endpoint is browser-consumed |\n| Background job \u002F Webhook | `workers\u002F**`,`jobs\u002F**`,`webhook*`,`cron*`,`*.worker.ts` | Trigger condition, retry, idempotency | A |\n| Configuration | `*.env*`,`*.yaml`,`*.toml`,`Dockerfile`,`docker-compose*`,`.github\u002Fworkflows\u002F**` | Env presence, build\u002Fruntime, CI green | A |\n| Docs only | `*.md`,`docs\u002F**` | Docs review checklist | A or B depending on audience of the doc |\n\n**Do not fabricate surfaces that do not exist in the diff.** An empty section tells the tester nothing and dilutes the real work.\n\n### 2c. Browser-observable effects (even in backend-only PRs)\n\nFile-pattern matching alone is not sufficient to decide the B section. A backend-only PR can still have browser-observable effects that only a real browser can verify — curl cannot reliably reproduce CORS preflight, cookie partitioning, `\u003Cscript>` tag loading, `Content-Type` sniffing, or downstream dashboard behaviour.\n\n**Include a B (QA\u002FBrowser) section even when no frontend files changed, if any of the following are true:**\n\n| Browser-observable effect | B-section scenario to include |\n|---|---|\n| Public endpoint consumed by a browser (`\u002Fconfig\u002F:id`, SDK loader, webhook receiver) | \"Fetch from DevTools Console\" scenario |\n| CORS \u002F preflight behaviour changed | Preflight-from-real-browser scenario (curl does not fully simulate preflight) |\n| `Content-Type` \u002F CSP \u002F cookie attributes changed | DevTools Network tab verification |\n| Backend APIs consumed by an **existing unchanged** frontend in the same repo | Consumer-renders-successfully scenario (regression-style) |\n| Script\u002Fasset delivered to end-user browsers (SDK, embed, widget) | \"Load from test HTML page\" scenario |\n\nThe test for whether to include a B section is not \"did a `.tsx` file change\" but **\"can a human verify this change by opening a browser?\"** When in doubt, include a minimal B section of 1–2 scenarios rather than omit it — a missing B section on a release that touches public endpoints is a common source of production surprises.\n\n## Step 3: Scenario Generation\n\nFor each (audience, surface) pair that exists, generate 2–5 scenarios. A scenario is one testable behaviour, not one changed file.\n\nEach scenario has:\n\n- **Goal** — one sentence\n- **Preconditions** — state the app must be in\n- **Steps** — numbered, imperative, specific enough for a tester who did not write the code\n- **Expected** — concrete values (`expect 3 rows with status = 'active'`), not \"it should work\"\n- **Checkbox** — one `- [ ]` at the top\n\n**Anti-pattern**: `- [ ] Test that the feature works`. That is not a scenario. If you cannot state *what to run\u002Fclick and what to see*, the scenario is not ready.\n\nScenario sources, in priority order:\n\n1. PR body \u002F task Acceptance Criteria (\"How to test\", \"動作確認\", \"Reviewer should verify\")\n2. Commit messages\n3. Surface-driven scenarios (from 2b)\n4. Diff edge cases (new `NOT NULL` → null-input scenario; new permission check → unauthorised-role scenario)\n\n### Triggered security negative paths\n\nRead the shared security omission lenses at `ywc-spec-validate\u002Freferences\u002Fsecurity-qa-lenses.md`, resolved from the skills root containing this installed skill (not the working directory), against the already-acquired diff, task\u002Fspec and verified project context. If the installed reference is missing or unreadable, return `NEEDS_CONTEXT` identifying that path and request installation of `ywc-spec-validate`; do not silently skip security checks or invent a replacement rubric. Add one negative-path scenario for each triggered lens, with concrete preconditions, steps and observable denial\u002Fsafe-failure outcome. Deduplicate overlapping scenarios without dropping any triggered check: record all covered lens names on the retained scenario and preserve each distinct expected outcome. These security scenarios are in addition to the 2–5 per (audience, surface) pair above, so that limit never forces a triggered lens to be dropped; they still count toward the Step 4 scenario count. Do not invent untriggered security surfaces; record N\u002FA reasons where appropriate. Scope manual checks to safe test fixtures and authorized environments; this does not authorize attacks or production probes.\n\nReuse evidence from the selected input mode. Do not fetch a PR body solely to import Impact; Impact already present in acquired evidence may inform scenarios but never substitutes for verified scope. Do not add unconditional PR-body acquisition, another diff read, screenshot operations or broad collection for this feature.\n\n## Step 4: Tier Decision\n\nCount scenarios after Step 3. Use the count to choose layout. This protects testsheets from becoming unread 1000-line walls.\n\n| Tier | Scenarios | Layout |\n|---|---|---|\n| **S** | ≤ 20 | Single file, A+B sections, **no TOC**, **no collapsible details** |\n| **M** | 21–40 | Single file, A+B sections, **auto-TOC** at top, **`\u003Cdetails>` collapsible** for Prerequisites + Edge Cases |\n| **L** | > 40 scenarios **OR** range > 50 commits | **Ask user** before writing:\u003Cbr>1. Single file with TOC + collapsible (as M)\u003Cbr>2. `--split` into `\u003Cslug>-dev.md` + `\u003Cslug>-qa.md`\u003Cbr>3. Split by Phase \u002F feature (user names each file)\u003Cbr>Proceed only after the user chooses. `--force-single` bypasses the prompt. |\n\nExplicit flags override the tier default -\n\n- `--split` → physical split, regardless of tier\n- `--audience dev|qa` → single-audience file, skip the other section entirely\n- `--force-single` → one file even if count > 40\n- `--no-toc` → suppress TOC even in M\u002FL\n\n## Step 5: Write the Testsheet\n\n### Output format\n\nDefault is a Markdown testsheet (`.md`) — keep this for the existing `-v\u003CN>`\nappend and front-matter sign-off workflow. With `--format html`, write an\n**interactive HTML testsheet** instead, following\n[html-output.md](..\u002Freferences\u002Fhtml-output.md): each scenario checkbox is\nclickable, sign-off state (tester name, Pass\u002FFail\u002FBlocked, notes) persists in\nthe browser via `localStorage`, and a `Copy as Markdown` button exports the\ncurrent sign-off state back to the Markdown testsheet format so the result can\nbe pasted into a PR or committed. HTML mode is recommended when the testsheet\nwill be handed to a QA\u002FPM who signs off in a browser; the filename follows the\nDefault filename table below with an `.html` extension. The Markdown surface\n(front matter included) is preserved inside the file.\n\n### Default filename\n\n| Input | Single file (default) | `--split` mode |\n|---|---|---|\n| PR | `pr-\u003Cnumber>-\u003Cslug>.md` | `pr-\u003Cnumber>-\u003Cslug>-dev.md` + `...-qa.md` |\n| Task | `task-\u003Cphase>-\u003Csequence>-\u003Cslug>.md` | `...-dev.md` + `...-qa.md` |\n| Task Range | `tasks-\u003Cstart-prefix>-\u003Cend-prefix>-\u003Cslug>.md` | `...-dev.md` + `...-qa.md` |\n| Range | `range-\u003Cshort-start>-\u003Cshort-end>-\u003Cslug>.md` (tag names used verbatim when both endpoints are tags, e.g. `range-v1.2-v1.3-\u003Cslug>.md`) | `...-dev.md` + `...-qa.md` |\n| Diff | `\u003Cyyyymmdd-HHMM>-\u003Cbranch-slug>.md` | `...-dev.md` + `...-qa.md` |\n\nIf the target file already exists, append `-v\u003CN>` (never overwrite). Existing files may carry a tester's partial sign-off.\n\n### Single-file template (default, A+B)\n\nWrite `\u003Coutput-dir>\u002F\u003Cfilename>.md` using this skeleton. Keep front matter keys, section numbers, file names, code snippets, and template structure in English regardless of `--lang`; they are machine surfaces. M\u002FL tier inserts a TOC block and wraps Prerequisites\u002FEdge Cases in `\u003Cdetails>`; S tier skips both.\n\n````markdown\n---\nsource: \u003Cpr#N | task:\u003Cname> | task-range:\u003Cstart>..\u003Cend> | diff | range>\npr_url: \u003Curl or \"\">\ntask_range_spec: \u003Ce.g. \"000012-010..000019-010\" or \"\">\ntask_count: \u003Cnumber or \"\">\nrange_spec: \u003Ce.g. \"main..feature-x\" or \"\">\nrange_start: \u003Cshort SHA or tag, or \"\">\nrange_end: \u003Cshort SHA or tag, or \"\">\nbranch: \u003Chead branch, or \"\" for ranges crossing branches>\nbase_branch: \u003Cbase>\nauthor: \u003Cusername>\ngenerated_at: \u003CISO 8601>\ntier: \u003CS | M | L>\ndev_tester: \"\"\ndev_status: pending\nqa_tester: \"\"\nqa_status: pending\n---\n\n# Test Case: \u003CTitle>\n\n\u003C!-- TOC block: M\u002FL tier only -->\n## Table of Contents\n- [1. Summary](#1-summary)\n- [2. Prerequisites](#2-prerequisites)\n- [A. Developer Verification](#a-developer-verification)\n  - [A.1 Database](#a1-database--table)\n  - [A.2 API](#a2-api)\n  - [A.3 Background Jobs](#a3-background-jobs--workers)\n  - [A.4 Configuration](#a4-configuration)\n  - [A.5 Dev Edge Cases](#a5-dev-edge-cases)\n  - [A.6 Dev Sign-off](#a6-dev-sign-off)\n- [B. QA \u002F Browser Verification](#b-qa--browser-verification)\n  - [B.1 UI \u002F Browser Scenarios](#b1-ui--browser-scenarios)\n  - [B.2 User-visible Edge Cases](#b2-user-visible-edge-cases)\n  - [B.3 Regression](#b3-regression)\n  - [B.4 QA Sign-off](#b4-qa-sign-off)\n\n## 1. Summary\n\n\u003C2–4 sentences explaining what changed and why a human needs to verify it.>\n\n## 2. Prerequisites\n\n\u003C!-- M\u002FL tier: wrap in \u003Cdetails> tag, open by default -->\n\n### 2.0 Common (both audiences)\n- [ ] …\n\n### 2.A Dev-only\n- [ ] …\n\n### 2.B QA-only\n- [ ] …\n\n## A. Developer Verification\n\n> **Audience**: Backend \u002F DB \u002F DevOps engineer  \n> **Tools**: psql, gh, curl, docker  \n> **Gate**: pre-merge\n\n### A.1 Database \u002F Table\n\u003Conly if surface exists>\n\n#### A.1.1 \u003Cscenario name>\n- [ ] **Goal**: …\n- **Preconditions**: …\n- **Steps**:\n  1. …\n- **Expected**: …\n\n### A.2 API\n### A.3 Background Jobs \u002F Workers\n### A.4 Configuration\n### A.5 Dev Edge Cases\n\n\u003C!-- M\u002FL tier: wrap A.5 in \u003Cdetails> -->\n\n- [ ] \u003Cboundary \u002F null \u002F error path>\n\n### A.6 Dev Sign-off\n- **Tester**:\n- **Date**:\n- **Result**: ☐ Pass ☐ Fail ☐ Blocked\n- **Notes**:\n\n## B. QA \u002F Browser Verification\n\n> **Audience**: QA \u002F PM \u002F Product Owner \u002F Designer  \n> **Tools**: Chrome + DevTools, admin UI, test origin  \n> **Gate**: pre-release\n\n### B.1 UI \u002F Browser Scenarios\n\u003Conly if a browser-observable surface exists>\n\n### B.2 User-visible Edge Cases\n\n\u003C!-- M\u002FL tier: wrap in \u003Cdetails> -->\n\n### B.3 Regression\n\u003Conly if --include-regression>\n\n### B.4 QA Sign-off\n- **Tester**:\n- **Date**:\n- **Result**: ☐ Pass ☐ Fail ☐ Blocked\n- **Notes**:\n\n## Appendix\n\u003Coptional — long SQL, payload samples, troubleshooting>\n````\n\n### Split-file variants (`--split` or `--audience`)\n\nEach file contains only that audience's Summary, Prerequisites, Scenarios, and Sign-off. Add a **Cross-reference** line at the bottom pointing to the counterpart (`Developer sheet: .\u002F\u003Cslug>-dev.md`, `QA sheet: .\u002F\u003Cslug>-qa.md`). Omit cross-reference when `--audience` was explicit and only one file is produced.\n\n## Length Management Guidelines\n\nApply these regardless of tier to keep testsheets lean:\n\n1. **Prerequisites common-prefix + audience-suffix** — shared rows live once in `## 2.0 Common`; dev-only \u002F qa-only rows live in `## 2.A` \u002F `## 2.B`. Never duplicate.\n2. **Extract long verification material** — any SQL \u002F script > 20 lines lives in `scripts\u002Fqa\u002F*.sql` (or similar) and the scenario references the path only. The scenario body stays short.\n3. **Regression by reference** — instead of repeating known-good regression flows, link to prior testsheets (e.g. `Stripe release testsheet: .\u002Fpr-230-stripe-release.md`).\n4. **Appendix at the bottom** — overflow content (detailed troubleshooting, sample payloads) goes under `## Appendix` and is linked from the relevant scenario.\n\nApplied consistently, these four rules keep most M-tier testsheets under ~800 lines and prevent L-tier by shifting volume out of the main flow.\n\n## Step 6: Validate and Report\n\nBefore finishing:\n\n1. File(s) exist and non-empty.\n2. Checkbox count ≥ 3.\n3. Every scenario has a concrete **Expected** result.\n4. If M\u002FL tier and `--no-toc` not set, TOC present at the top with anchors matching headings.\n5. If split layout, cross-reference line present in each file.\n\nReport:\n\n```text\nTestsheet generated: \u003Cabsolute path(s)>\nSource: \u003Cpr#N | task:\u003Cname> | task range:\u003Ctask-start>..\u003Ctask-end> | diff | git range:\u003Cstart>..\u003Cend>>\nTier: \u003CS | M | L>\nLayout: \u003Csingle | split>\nAudiences: \u003CA only | B only | A+B>\nScenarios by surface: UI (3), DB (1), API (2), Jobs (4), Config (2)\nScenarios by audience: Dev \u003Cn>, QA \u003Cm>\nTotal: \u003Cscenarios> scenarios, \u003Ccheckboxes> checkboxes\nLanguage: \u003Cdetected-or-specified>\nNext steps:\n  - Dev signs off on Section A (pre-merge gate)\n  - QA signs off on Section B (pre-release gate)\n  - Update front-matter `dev_status:` \u002F `qa_status:` as work progresses\n```\n\n## Language Detection\n\nWhen `--lang` is not specified, resolve testsheet prose language using the\nshared Codex YWC policy:\n[`..\u002Freferences\u002Flanguage-resolution.md`](..\u002Freferences\u002Flanguage-resolution.md).\nThe order is explicit `--lang` > project `.codex\u002Fywc.json` > project guidance\n(`AGENTS.md`, `CODEX.md`, `CLAUDE.md`) > user `~\u002F.codex\u002Fywc.json` > ask user.\n\nRecent testsheets in the output directory and the project `README.md` may be\nused only as non-authoritative context when asking the user. They do not create\na final fallback default.\n\nSupported language codes are `ja`, `ko`, `en`, `zh`, and `es`. Treat aliases such as `japanese`, `korean`, `english`, `chinese`, `Chinese (Simplified)`, `中文`, `spanish`, `espanol`, and `español` as requests for the matching code; `zh` \u002F `chinese` means Simplified Chinese unless the user explicitly asks for a different Chinese locale.\n\nYAML front matter keys, section numbers, file names, code snippets, command names, and the template skeleton stay in English regardless of `--lang`. Only prose (Summary, Goal, Steps, Expected, Notes, Edge Cases, and audience-facing test descriptions) follows the chosen language. Technical terms that are part of the testsheet contract or tooling surface, such as Developer Verification, QA \u002F Browser, PR, Task, Git Range, Expected, front matter, Status, and Sign-off, stay English in `zh` and `es` prose.\n\n## Dry Run Mode\n\nWith `--dry-run`, perform Steps 1–4 as usual, then print:\n\n- Resolved input mode and source\n- Tier decision (S\u002FM\u002FL) and scenario count\n- Target filename(s) and full output path(s)\n- Surface classification (which sections will exist)\n- Scenarios per (audience, surface) pair\n- Detected language\n\nDo **not** write any file. Exit after printing.\n\n## Error Handling\n\n| Situation | Behaviour |\n|---|---|\n| No input given | Stop, ask for PR \u002F task \u002F range \u002F `--from-diff` |\n| Multiple inputs given (PR + Range, Task + Range, Range + `--from-diff`, etc.) | Stop, ask which mode |\n| Three-dot range `A...B` passed | Reject and ask for the two-dot `A..B` form instead |\n| Task range endpoint missing or ambiguous | Report the endpoint and similar task names; stop without trying `git rev-parse` |\n| Task range start appears after end | Stop and ask whether the user intended the reverse range |\n| Range ref invalid (`git rev-parse` fails on `\u003Cstart>` or `\u003Cend>`) | Report which ref failed; stop |\n| Range empty (`\u003Cend>` has no commits absent from `\u003Cstart>`) | Report \"range is empty\"; stop |\n| Range > 50 commits | Escalate to L-tier handling (ask user how to proceed) |\n| Range HEAD belongs to an open\u002Fmerged PR (≥ 80% overlap) | Stop and ask whether to switch to PR mode |\n| Range has ≥ 70% low-quality commit headlines | Warn and ask whether to proceed |\n| `--split` and `--force-single` both set | Stop, ask which is intended |\n| `gh` not authenticated (PR input or range PR auto-suggest) | Ask user to run `gh auth login`; stop |\n| PR not found | Report the PR number and stop |\n| Task not found | List similar tasks (fuzzy match) and stop |\n| Empty diff (diff input) | Report \"nothing to test\" and stop |\n| Output directory not writable | Report failing path and stop (no silent fallback) |\n| Target file already exists | Append `-v\u003CN>` suffix; report new filename |\n| L tier detected and no explicit `--split`\u002F`--force-single` | Ask user how to proceed before writing |\n\n## Integration\n\nDownstream of the implementation-oriented `ywc` skills:\n\n- **ywc-sequential-executor** — generates the PR\u002FTask this skill tests\n- **ywc-parallel-executor** — same, for parallel execution\n- **ywc-merge-dependabot** — merged dependency bumps that often warrant a smoke testsheet\n\nInvoke this skill after those finish, when a human tester is ready to validate.\n\n## Examples\n\nFor invocation examples across PR, task, task range, diff, and git range modes, read [references\u002Fexamples.md](references\u002Fexamples.md).\n\n## Output Format\n\nInclude `Status: \u003CDONE | DONE_WITH_CONCERNS | BLOCKED | NEEDS_CONTEXT>`, the generated testsheet path or dry-run summary, detected scope, and any blocker.\n\n## Validation\n\nBefore finalizing, verify that input mode is unambiguous, the diff\u002Frange is non-empty, output path handling is deterministic, and no file is written during `--dry-run`.\n",{"name":5,"description":7},"main","7e8ac872e53df8c89d75663e031626281a2a2f5c","plugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase",1,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit","https:\u002F\u002Fgithub.com\u002Fyongwoon","https:\u002F\u002Fwww.webhani.com","2026-11-08 17:14:59",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add yongwoon\u002Fywc-agent-toolkit --skill ywc-gen-testcase",{"tool":37,"command":38},"bunx","bunx skills add yongwoon\u002Fywc-agent-toolkit --skill ywc-gen-testcase",{"tool":40,"command":41},"pnpm","pnpm dlx skills add yongwoon\u002Fywc-agent-toolkit --skill ywc-gen-testcase","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FSKILL.md",true,[45,51,55,59,63,67,71,75,79],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"agents\u002Fopenai.yaml","REFERENCE","参考文档",281,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002Fagents\u002Fopenai.yaml",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"evals\u002Fevals.json",2085,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002Fevals\u002Fevals.json",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"README.en.md",11444,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FREADME.en.md",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"README.es.md",13408,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FREADME.es.md",{"path":64,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":65,"sourceHtmlUrl":66},"README.ja.md",11982,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FREADME.ja.md",{"path":68,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":69,"sourceHtmlUrl":70},"README.ko.md",14457,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FREADME.ko.md",{"path":72,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":73,"sourceHtmlUrl":74},"README.md",12231,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FREADME.md",{"path":76,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":77,"sourceHtmlUrl":78},"README.zh.md",11379,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002FREADME.zh.md",{"path":80,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":81,"sourceHtmlUrl":82},"references\u002Fexamples.md",1355,"https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-gen-testcase\u002Freferences\u002Fexamples.md",[84,90,96,103,110,117],{"skillNo":85,"name":86,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":18},"01M72TQ84GTCPCV2VJ778V17TW","missing-frontmatter","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fcodex-description-limit\u002Fmissing-frontmatter\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":18},"01M72TQ6M4BFWE7P9EPW0GZDCM","malformed-frontmatter","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fcodex-description-limit\u002Fmalformed-frontmatter\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":18},"01M72TQ3ESGKXZ4QRWGTX92BKJ","fixture-501-fail","(ywc) Use when aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa Do not use for x.","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fcodex-description-limit\u002F501-fail\u002FSKILL.md",[],[],{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":109,"publishedAt":18},"01M72TQ0HJ8H3R37J9WPWGKJZR","fixture-500-pass","(ywc) Use when aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa Do not use for x.","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fcodex-description-limit\u002F500-pass\u002FSKILL.md",[],[],{"skillNo":111,"name":112,"nameZh":6,"description":113,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":114,"categories":115,"keywords":116,"publishedAt":18},"01M72TPWNCC7NEYD0D9N12KSQS","ywc-worktrees","(ywc) Use when creating, auditing, pruning, or resolving git worktrees for isolated task workspaces, whether delegated from an executor or used standalone. Triggers: \"worktree 생성\", \"worktree audit\", \"worktree 정리\", \"worktree prune\", \"ywc-worktrees\", \"create worktree\". Do not use for branch creation\u002Fdeletion alone, PR creation (use ywc-create-pr), CI or merge delivery (use ywc-finish-branch), task orchestration (use ywc-sequential-executor or ywc-parallel-executor), or non-git isolation.","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-worktrees\u002FSKILL.md",[],[],{"skillNo":118,"name":119,"nameZh":6,"description":120,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":121,"categories":122,"keywords":123,"publishedAt":18},"01M72TPRNR3E6QWB91GM98ABBV","ywc-wayfinder","(ywc) Use when a large or uncertain change needs multi-session discovery with a deterministic local map and exactly one active ticket before planning or implementation. Triggers: \"wayfinder\", \"길찾기\", \"discovery map\", \"multi-session discovery\", \"탐색 ticket\", \"経路整理\", \"探索マップ\", \"ywc-wayfinder\". Do not use for ordinary planning (use ywc-plan), idea clarification (use ywc-brainstorm), direct implementation, or tracker\u002Fdatabase writes outside the local Markdown map.","https:\u002F\u002Fgithub.com\u002Fyongwoon\u002Fywc-agent-toolkit\u002Fblob\u002Fmain\u002Fplugins\u002Fywc-agent-toolkit\u002Fskills\u002Fywc-wayfinder\u002FSKILL.md",[],[]]