feature-workflow
/feature-workflow
Prowadzi funkcję od zapytania do ostatniego commitu: specyfikacja, plan, testy najpierw, sprawdzenie w przeglądarce i testy end-to-end.
- Autor
- dvasyliev ↗
- Repozytorium
- frontend-harness ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 11.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 11.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Prowadzi funkcję od zapytania do ostatniego commitu: specyfikacja, plan, testy najpierw, sprawdzenie w przeglądarce i testy end-to-end.
Oryginalny opis (EN)
Use when building any user-facing feature, from the request to the last commit - spec, plan, tests first, browser check, end-to-end tests and improving the harness.
Jak działa
- Spec
- Plan
- Unit tests first (with unit)
- Browser check (with e2e)
- End-to-end tests (with e2e)
- Improve the harness
Podgląd SKILL.md
--- name: feature-workflow description: Use when building any user-facing feature, from the request to the last commit - spec, plan, tests first, browser check, end-to-end tests and improving the harness. --- # Building a feature Go through the steps below in order. A step or sentence marked with a module (such as "With `e2e`") applies only when that module is in the harness modules listed in `.claude/rules/harness.md`; skip it otherwise. ## Spec Start from the feature's spec in `docs/superpowers/specs/`, if it has one. Note what existing code is reused and what the feature replaces. Specs and plans go where superpowers puts them, in the docs language given in `.claude/rules/harness.md`: a spec (from `superpowers:brainstorming` or written by hand) in `docs/superpowers/specs/YYYY-MM-DD-<topic>-design.md`, a plan in `docs/superpowers/plans/YYYY-MM-DD-<feature-name>.md` with the `writing-plans` header and a `**Spec:**` line pointing to its spec. Then decide what the request needs: - **Intent unclear** (vague or garbled request, several readings): ask with `AskUserQuestion` before anything else, as many questions as it takes, with a recommended option first. Never ask what the code or existing patterns already answer. - **Intent clear, but a real design choice is open:** use `superpowers:brainstorming`. - **Intent and design both clear:** go straight to the plan. ## Plan Use `superpowers:writing-plans`. List each scenario's user steps with the expected outcome on `- expect:` lines. If `superpowers:brainstorming` produced a written spec, show the plan and wait for the user's approval. Otherwise start at once. Either way, carry it out as `.claude/rules/plan-execution.md` says. With `e2e`: make the end-to-end tests the plan's last task, so the final review that both ways of carrying it out end with covers them; that review is the feature's review. The browser check stays with you, not the subagents. Ask the user only about what the spec leaves open, not about how to execute. ## Unit tests first (with `unit`) Use `superpowers:test-driven-development`: write the tests, watch them fail with `npm test`, then implement. ## Browser check (with `e2e`) With the `playwright-cli` skill, drive the feature on the running dev server and confirm the `- expect:` outcomes. `@playwright/cli` is a devDependency of the project: run every command as `npx playwright-cli <command>`, where the skill writes `playwright-cli <command>`. Never install it globally, even when the skill says to. Open a browser of your own, named after the feature (`npx playwright-cli -s=<feature> open`), and pass the same `-s` to every command; the default session is shared with every other agent and with the user. Close it when done (`npx playwright-cli -s=<feature> close`). ### Screenshots Every screenshot stays in the conversation and is re-read on every later turn, so take few and small ones. - Check behaviour with `npx playwright-cli snapshot` (text), not screenshots. Search it instead of reading it whole when you look for one thing. - Take a screenshot only to judge how something looks: layout, colours, spacing. - Screenshot the element in question (`npx playwright-cli screenshot e5`), not the whole page. Never use `--hires`. - Fix the window size before screenshots: `npx playwright-cli resize 1280 800` for desktop, `npx playwright-cli open --mobile` for phone. - One screenshot per state you need to judge. Do not re-take one to confirm a change a snapshot already shows. ## End-to-end tests (with `e2e`) Write them for the feature's happy paths only, one per main scenario. While working, run only the feature's own spec files (`npx playwright test tests/x/y.spec.ts`). Commits after each task are fine; run the whole suite with `npm run verify` once, before the last commit. After every run, read its log for errors as `tests-e2e.md` says; a run that passes but logs an error the feature caused is not done. ## Improve the harness Invoke the `improve-harness` skill.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · dvasyliev/frontend-harness | 100 | MIT |
| siva01c/claude-plugins | 100 | MIT |
| arraydude/agent-skills | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 80138db320de (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100workflow-build-feature
Buduje funkcję od początku do końca: specyfikacja i TDD, implementacja, testy jednostkowe, kontrola Playwright i pull request. - 100test-plan
Planuje testy funkcji: mapuje kryteria akceptacji na testy, szuka luk i uruchamia główne przepływy end-to-end. - 100v1-e2e-testing
Pisze i naprawia testy end-to-end w przeglądarce, zwłaszcza w Playwright, i diagnozuje niestabilne testy. - 90app-feature-craft
Prowadzi implementację lub ulepszanie przepływu użytkownika w aplikacji: UX, frontend, API, stany zadań AI, testy i weryfikacja end-to-end do gotowej funkcji. - 96testing-workflow
Pisze, uruchamia i debuguje testy na każdym poziomie, od jednostkowych po end-to-end. - 100test-browser
Uruchamia testy end-to-end w przeglądarce na stronach zmienionych w PR: mapuje pliki na trasy i zapisuje zrzuty ekranu.