
implement
by ArvorCo
implementは、other分野における実用的なスキルです。複雑な課題への対応力を強化し、業務効率と成果の質を改善します。
SKILL.md
name: implement description: "Execute implementation workflow phase by phase. Use after analysis passes. Triggers on: start implementation, implement feature, begin coding."
Implementation Workflow Executor
Guide systematic implementation of features using TDD and quality-first approach.
Note: This skill uses generic placeholders. Adapt commands and paths to your project:
- Quality commands: Check your
package.json,Makefile, or build config- File paths: Adjust based on your project structure
- Examples show common conventions - your project may differ
SpecKit Workflow
This skill is Step 6 of 6 in the Relentless workflow:
specify → plan → tasks → convert → analyze → implement
Prerequisites:
spec.md- Feature specificationplan.md- Technical implementation plantasks.md- User stories with acceptance criteriaprd.json- Converted PRD with routing metadatachecklist.md- Quality validation checklist- Analysis must PASS (run
/relentless.analyzefirst)
The Job
Implement user stories one at a time, following strict TDD and updating all tracking files.
Before You Start
- Read the Constitution - Review your project's constitution (if exists) for governance principles
- Read prompt.md - Review workflow guidelines (if exists)
- Read progress.txt - Check learnings from previous iterations
- Review artifacts - Ensure spec.md, plan.md, tasks.md, prd.json, checklist.md exist
- Verify Analysis Passed - Run
/relentless.analyzeif not already done - Identify Quality Commands - Find your project's typecheck, lint, and test commands
- Check Branch - Verify you're on the correct branch from PRD
branchName
CRITICAL: Quality Gates (Non-Negotiable)
Before marking ANY story as complete, ALL quality checks must pass.
Find your project's commands (examples for different ecosystems):
# JavaScript/TypeScript (npm/yarn/bun/pnpm):
npm run typecheck && npm run lint && npm test
# Python:
mypy . && ruff check . && pytest
# Go:
go build ./... && golangci-lint run && go test ./...
# Rust:
cargo check && cargo clippy && cargo test
Requirements:
- Typecheck: 0 errors
- Lint: 0 errors AND 0 warnings
- Tests: All pass
If ANY check fails, DO NOT mark the story as complete. Fix the issues first.
TDD Workflow (MANDATORY)
For EVERY story, follow strict Test-Driven Development:
Step 1: Write Failing Tests First (RED)
# Create test file if needed
# Write tests that define expected behavior
# Run your test command - tests MUST fail initially
Step 2: Implement Minimum Code (GREEN)
# Write only enough code to pass tests
# Run tests - they MUST pass now
Step 3: Refactor
# Clean up while keeping tests green
# Run tests - they MUST still pass
Do NOT skip TDD. Tests are contracts that validate your implementation.
Research Phase (if story has research: true)
If the current story has research: true in prd.json and no research file exists yet:
- Explore the codebase - Find relevant files, patterns, and dependencies
- Document findings in
relentless/features/<feature>/research/<story-id>.md:- Existing patterns that should be followed
- Files that will likely need modification
- Dependencies and integration points
- Potential gotchas or edge cases
- Recommended implementation approach
- Do NOT implement - only research and document
- Save your findings to the research file and end your turn
Per-Story Implementation Flow
For each story (in dependency order):
1. Identify the Story
- Read
prd.jsonto find the next story wherepasses: false - Check dependencies are met (dependent stories have
passes: true) - Read the story's acceptance criteria
- Check routing metadata (complexity, model assigned)
2. Find Relevant Checklist Items
- Open
checklist.md - Find items tagged with
[US-XXX]for this story - Note any governance/compliance items
- Ensure Quality Gates and TDD items are included
3. Implement with TDD
Follow the TDD workflow above for each acceptance criterion.
4. Update tasks.md
As you complete each criterion:
# Change from:
- [ ] Criterion text
# To:
- [x] Criterion text
5. Update checklist.md
For each verified checklist item:
# Change from:
- [ ] CHK-XXX [US-001] Description
# To:
- [x] CHK-XXX [US-001] Description
6. Run Quality Checks
# Run your project's quality commands
# All must pass with 0 errors/warnings
7. Commit Changes
git add -A
git commit -m "feat: US-XXX - Story Title"
8. Update prd.json
Set the story's passes field to true:
{
"id": "US-001",
"title": "...",
"passes": true, // <- Change from false to true
...
}
9. Update progress.txt
Append progress entry:
## [Date] - US-XXX: Story Title
**Implemented:**
- What was built
- Key decisions made
**Files Changed:**
- path/to/file (new/modified)
**Tests Added:**
- path/to/test.file
**Learnings:**
- Patterns discovered
- Gotchas encountered
**Constitution Compliance:**
- [list principles followed]
---
Check for Queued Prompts
Between iterations, check .queue.txt for user input:
# If .queue.txt exists, read and process it
# Acknowledge in progress.txt
# Process in FIFO order
File Update Summary
After completing each story, these files MUST be updated:
| File | Update |
|---|---|
tasks.md | Check off - [x] completed acceptance criteria |
checklist.md | Check off - [x] verified checklist items |
prd.json | Set "passes": true for the story |
progress.txt | Append progress entry with learnings |
Implementation Phases
Phase 0: Setup
- Infrastructure, tooling, configuration
- Usually US-001 type stories
Phase 1: Foundation
- Data models, types, schemas
- Base utilities and helpers
- Core infrastructure
Phase 2: User Stories
- Feature implementation
- Follow dependency order strictly
Phase 3: Polish
- E2E tests
- Documentation
- Performance optimization
Stop Condition
After completing a user story, check if ALL stories have passes: true.
If ALL stories are complete and passing, output:
<promise>COMPLETE</promise>
If there are still stories with passes: false, end your response normally (another iteration will pick up the next story).
Common Pitfalls to Avoid
- Skipping TDD - Never implement without tests first
- Suppressing lints - Fix issues properly, don't disable rules
- Large commits - Keep commits focused and atomic
- Missing typecheck - Always run typecheck before commit
- Ignoring progress.txt - Read learnings from previous iterations
- Not checking queue - Always check
.queue.txtfor user input - Skipping analysis - Run
/relentless.analyzebefore implementing - Ignoring routing metadata - Check story complexity and model assignment
Notes
- Work on ONE story at a time
- Follow dependency order strictly
- Never skip TDD - tests come FIRST
- Never skip quality checks
- Commit after each story
- Update ALL tracking files
- Check
.queue.txtfor mid-run input - This is a guided workflow for systematic implementation
- Adapt all commands and paths to your project's specific setup
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