ワンクリックで
dev
Pick the highest priority task from the sprint backlog and implement it with TDD, security scanning, and browser-based E2E testing
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。
メニュー
Pick the highest priority task from the sprint backlog and implement it with TDD, security scanning, and browser-based E2E testing
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
| name | dev |
| description | Pick the highest priority task from the sprint backlog and implement it with TDD, security scanning, and browser-based E2E testing |
| user-invocable | true |
Senior software engineer workflow: pick the highest-priority uncompleted task from the current sprint backlog and implement it with test-driven development, integrated security scanning, and Playwright E2E testing.
One task per /dev invocation. Do not combine tasks.
Look for the latest sprints/vN/TASKS.md file. Read it and identify the highest-priority uncompleted task (first - [ ] item, preferring P0 over P1 over P2).
PRD.md for architecture and requirementsWALKTHROUGH.mdBefore writing any implementation code, write tests based on the task type:
For logic/utility tasks — Write unit tests:
npx vitest run [test-file]
For API route tasks — Write integration tests:
# Test the API endpoint with expected inputs/outputs
For UI/page tasks — Write Playwright E2E tests:
# Install Playwright if not already installed
npx playwright install chromium
# Write tests that:
# 1. Navigate to the page
# 2. Interact with elements (click, type, select)
# 3. Take screenshots at key steps
# 4. Assert visible elements and text
Save to tests/screenshots/taskN-stepN-description.png. Take screenshots BEFORE and AFTER key interactions. If a test fails, read the screenshot to debug visually.
tests/
screenshots/
task3-01-login-page.png
task3-02-after-login.png
task3-03-dashboard-loaded.png
unit/
auth.test.ts
metrics.test.ts
integration/
api-auth.test.ts
api-metrics.test.ts
e2e/
auth-flow.spec.ts
dashboard.spec.ts
Write the minimum code needed to make the tests pass:
data-testid attributes to interactive elements (for Playwright)# Run the specific test
npx vitest run tests/[file]
# or
npx playwright test tests/[file]
If tests fail:
After tests pass, run security scanning:
# Static analysis
npx semgrep --config auto src/ --quiet
# Dependency audit
npm audit
If findings exist:
Mark the task as complete:
- [x] Task N: [description] (P0)
- Acceptance: [criteria]
- Files: [files]
- Completed: [date] -- [brief note of what was done]
Create a git commit with a clear message:
git add -A
git commit -m "feat(vN): Task N -- [description]
- Implemented [what]
- Tests: [N unit, N integration, N e2e]
- Security: semgrep clean, npm audit clean"
/dev invocation. Don't combine tasks.data-testid attributes for Playwright selectors (not CSS classes).Read TASKS.md
|
Pick highest priority task
|
Write tests FIRST (TDD)
|
Run tests (should fail) --- FAIL ---> Fix test
|
PASS (tests are correct)
|
Implement the code
|
Run tests --- FAIL ---> Fix code, re-run
|
PASS
|
semgrep scan / npm audit --- FINDINGS ---> Fix, re-run
|
CLEAN
|
Update TASKS.md + Git commit
Interact with Basecamp via the Basecamp CLI. Full API coverage: projects, todos, cards, messages, files, schedule, check-ins, timeline, recordings, templates, webhooks, subscriptions, lineup, chat, gauges, assignments, notifications, and accounts. Use for ANY Basecamp question or action.
Brainstorm requirements and create sprint PRDs with atomic tasks. Creates sprints/vN/PRD.md and TASKS.md with user stories, architecture, and ordered task breakdowns.
Structured review for design system compliance, TypeScript quality, page structure, security, and accessibility
Multi-agent logo design pipeline orchestrating 6 agents through Strategy, Concepts, Figma, Code, Preview, and QA phases
Scaffold a new page following all Colaberry AI platform standards including locked theming and page structure
Safe refactoring workflow that reads before writing, plans before editing, and verifies after each change