
aico-backend-implement
by yellinzero
Build your AI team in seconds, start working immediately
SKILL.md
name: aico-backend-implement description: | Execute backend implementation with TDD (Test-Driven Development). Write failing test first, then implement, verify each step.
Use this skill when:
- User asks to "implement this backend task", "implement backend"
- Have an implementation plan ready and need to execute it
- Executing steps from /backend.plan output
- User says "start coding backend", "write backend code"
- User asks to "use TDD", "write test first", "test-driven" for backend code
- Fixing backend bugs (write failing test that reproduces bug first)
TDD Iron Law: NO PRODUCTION CODE WITHOUT A FAILING TEST FIRST TDD Cycle: RED (write failing test) → Verify fails → GREEN (minimal code) → Verify passes → REFACTOR
Prerequisites:
- MUST have task file in docs/reference/backend/tasks/ (use /backend.tasks first if not exists)
- MUST read constraints.md before writing any code. Flow: Check Task File → Read Constraints → TDD Cycle (RED→GREEN→REFACTOR) → Verify Each → Run All Tests → Update Task Status → Notify PM
Implement
Language Configuration
Before generating any content, check aico.json in project root for language field to determine the output language. If not set, default to English.
Process
-
Check task file EXISTS (MANDATORY):
- Look for
docs/reference/backend/tasks/{story-id}.md - If NOT exists → STOP and tell user: "Please run /backend.tasks first to break down the story"
- If exists → proceed to step 1
- Look for
-
Read constraints FIRST: Load
docs/reference/backend/constraints.md -
Execute each step in order:
- If test step: write failing test, verify it fails
- If impl step: write code, verify test passes
- Run verification command
- If fail → fix before proceeding
-
After all steps:
- Run full test suite
- Run type check
- Run linter
-
Update task status in
docs/reference/backend/tasks/{story}.md -
Notify PM for acceptance - Notify PM to verify (PM checks all related tasks and updates Story status)
-
Report completion (user decides when to commit)
Execution Flow
Check Task File → Read Constraints → Execute Steps (TDD) → Verify Each → Run All Tests → Mark Complete → Notify PM
TDD Integration
Step N: Write failing test
↓ Verify: test fails ❌
Step N+1: Implement code
↓ Verify: test passes ✅
Step Execution Rules
Rule 1: Follow Constraints
// ❌ Wrong: Ignore constraints
const password = user.password; // plaintext
// ✅ Right: Follow security constraints
const passwordHash = await bcrypt.hash(password, 10);
Rule 2: Verify Before Proceeding
Each step has a Verify section - MUST run it and confirm expected output.
Rule 3: Test First (TDD)
// ❌ Wrong order
Step 1: Implement createUser method
Step 2: Write test for createUser
// ✅ Correct order
Step 1: Write failing test for createUser
Step 2: Implement createUser method
Rule 4: No Skipping
- Execute ALL steps in order
- Do NOT combine steps
- Do NOT skip verification
Post-Implementation Checklist
- Run full test suite:
npm test - Run type check:
npx tsc --noEmit - Run linter:
npm run lint - Update task status in tasks file
- Notify PM: Tell PM backend tasks are complete, request acceptance
- Report completion - user decides when to commit
Error Handling
| Error Type | Action |
|---|---|
| Test failure | Debug, fix, re-run |
| Type error | Fix types, re-verify |
| Lint error | Fix style issues |
Key Rules
- ALWAYS read constraints before writing any code
- MUST follow TDD: test first, then implementation
- ALWAYS run verification for each step
- MUST run full test suite before marking complete
- MUST notify PM after completing all tasks (PM handles story verification)
- Let user decide when to commit
Common Mistakes
- ❌ Implement without task file → ✅ ALWAYS run /backend.tasks first to break down story
- ❌ Skip reading constraints → ✅ ALWAYS read before coding
- ❌ Write code before test → ✅ TDD: test first
- ❌ Skip verification → ✅ Run verify command for each step
- ❌ Skip full test suite → ✅ Run all tests before complete
- ❌ Directly update story status → ✅ Notify PM, let PM verify and update story
TDD Deep Dive
The TDD Iron Law
NO PRODUCTION CODE WITHOUT A FAILING TEST FIRST
Write code before the test? Delete it. Start over.
Red-Green-Refactor Cycle
RED → Verify Fails → GREEN → Verify Passes → REFACTOR → Repeat
1. RED - Write Failing Test
Write ONE minimal test showing expected behavior:
test('createUser hashes password before storing', async () => {
const user = await userService.createUser({
email: 'test@example.com',
password: 'plaintext123',
});
expect(user.passwordHash).not.toBe('plaintext123');
expect(await bcrypt.compare('plaintext123', user.passwordHash)).toBe(true);
});
2. Verify RED - Watch It Fail
MANDATORY. Never skip.
npm test path/to/test.ts
3. GREEN - Write Minimal Code
Write simplest code to pass the test. Don't add features not in test.
4. Verify GREEN - Watch It Pass
npm test path/to/test.ts
5. REFACTOR - Clean Up
Only after green. Keep tests passing.
Test Types
| Type | Purpose | Example |
|---|---|---|
| Unit | Single function | authService.validatePassword() |
| Integration | With real DB | userRepository.create() |
| API | HTTP endpoints | POST /auth/register |
TDD Red Flags - STOP and Start Over
- Code before test
- Test passes immediately
- Can't explain why test failed
- Rationalizing "just this once"
Iron Law
NO CODE WITHOUT APPROVED PLAN
This rule is non-negotiable. Before writing code:
- Task breakdown must exist and be approved
- Acceptance criteria must be defined
- Dependencies must be identified and available
Rationalization Defense
| Excuse | Reality |
|---|---|
| "It's a simple change" | Simple changes often have hidden complexity |
| "I'll document after coding" | Post-hoc documentation is always incomplete |
| "Tests can wait until later" | Untested code is broken code |
| "I know what needs to be done" | Assumptions without validation cause bugs |
スコア
総合スコア
リポジトリの品質指標に基づく評価
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
レビュー
レビュー機能は近日公開予定です