Systematic development workflow: Analyze → Plan → Execute → eLicit → eXamine. Use for ANY development task: features, bug fixes, refactoring, hotfixes. Triggers: "implement", "create", "build", "fix", "add feature", "refactor", "develop". Auto-detects project type (Laravel, Next.js, React, Swift) and loads framework-specific references. Enforces: files <100 lines, interfaces separated, SOLID principles, expert self-review, sniper validation. Modes: --auto (default), --manual, --skip-elicit
From fuse-ai-pilotnpx claudepluginhub fusengine/agents --plugin fuse-ai-pilotThis skill uses the workspace's default tool permissions.
references/00-init-branch.mdreferences/01-analyze-code.mdreferences/02-features-plan.mdreferences/03-execution.mdreferences/03.5-elicit.mdreferences/04-validation.mdreferences/05-review.mdreferences/06-fix-issue.mdreferences/07-add-test.mdreferences/08-check-test.mdreferences/09-create-pr.mdreferences/design/00-init-branch.mdreferences/design/01-analyze-design.mdreferences/design/02-search-inspiration.mdreferences/design/03-plan-component.mdreferences/design/04-code-component.mdreferences/design/05-add-motion.mdreferences/design/06-validate-a11y.mdreferences/design/07-review-design.mdreferences/design/08-sniper-check.mdEnables AI agents to execute x402 payments with per-task budgets, spending controls, and non-custodial wallets via MCP tools. Use when agents pay for APIs, services, or other agents.
Compares coding agents like Claude Code and Aider on custom YAML-defined codebase tasks using git worktrees, measuring pass rate, cost, time, and consistency.
Designs and optimizes AI agent action spaces, tool definitions, observation formats, error recovery, and context for higher task completion rates.
Current Task: $ARGUMENTS
Analyze → Plan → Execute → eLicit → eXamine
Complete development workflow for features, fixes, and refactoring.
BEFORE anything else, run this command to initialize APEX tracking:
mkdir -p .claude/apex/docs && cat > .claude/apex/task.json << 'INITEOF'
{
"current_task": "1",
"created_at": "'"$(date -u +"%Y-%m-%dT%H:%M:%SZ")"'",
"tasks": {
"1": {
"status": "in_progress",
"started_at": "'"$(date -u +"%Y-%m-%dT%H:%M:%SZ")"'",
"doc_consulted": {}
}
}
}
INITEOF
echo "✅ APEX tracking initialized in $(pwd)/.claude/apex/"
This creates:
.claude/apex/task.json - Tracks documentation consultation status.claude/apex/docs/ - Stores consulted documentation summariesThe PreToolUse hooks will BLOCK Write/Edit until documentation is consulted.
┌─────────────────────────────────────────────────────────────────┐
│ APEX WORKFLOW │
├─────────────────────────────────────────────────────────────────┤
│ 00-init-branch → Create feature branch │
│ 00.5-brainstorm → Design-first questioning (B) ← NEW │
│ 01-analyze-code → Understand codebase (A) │
│ 02-features-plan → Plan implementation (P) │
│ 03-execution → Write code with TDD (E) ← UPDATED │
│ 03.5-elicit → Expert self-review (L) │
│ 03.7-verification → Functional resolution check (V) ← NEW │
│ 04-validation → Verify quality (X) │
│ 05-review → Self-review │
│ 06-fix-issue → Handle issues │
│ 07-add-test → Write tests (TDD cycle) │
│ 08-check-test → Run tests │
│ 09-create-pr → Create Pull Request │
└─────────────────────────────────────────────────────────────────┘
| Phase | Skill | Invocation |
|---|---|---|
| 00.5 | brainstorming | Questions → alternatives → design doc → approval |
| 03 | tdd | RED (test) → GREEN (code) → REFACTOR cycle |
| 03.7 | verification | Re-read request → check criteria → confirm resolution |
| Phase | File | Purpose |
|---|---|---|
| 00 | references/00-init-branch.md | Create feature branch |
| 01 | references/01-analyze-code.md | Explore + Research (APEX A) |
| 02 | references/02-features-plan.md | TaskCreate planning (APEX P) |
| 03 | references/03-execution.md | Implementation (APEX E) |
| 03.5 | references/03.5-elicit.md | Expert self-review (APEX L) ← NEW |
| 04 | references/04-validation.md | sniper validation (APEX X) |
| 05 | references/05-review.md | Self-review checklist |
| 06 | references/06-fix-issue.md | Fix validation/review issues |
| 07 | references/07-add-test.md | Write unit/integration tests |
| 08 | references/08-check-test.md | Run and verify tests |
| 09 | references/09-create-pr.md | Create and merge PR |
1. 00-init-branch → git checkout -b feature/xxx
2. 00.5-brainstorm → Ask questions, propose alternatives, get design approval
3. 01-analyze-code → explore-codebase + research-expert
4. 02-features-plan → TaskCreate task breakdown
5. 03-execution → TDD: write test FIRST, then implement (files <100 lines)
6. 03.5-elicit → Expert self-review (elicitation techniques)
7. 03.7-verification → Verify functional resolution against original request
8. 04-validation → sniper agent (code quality)
9. 05-review → Self-review
10. 09-create-pr → gh pr create
1. 00-init-branch → git checkout -b fix/xxx
2. 01-analyze-code → Understand bug context
3. 07-add-test → TDD: write failing test FIRST (RED)
4. 03-execution → Fix the bug (GREEN)
5. 08-check-test → Verify test passes + no regressions
6. 03.7-verification → Verify original bug is functionally resolved
7. 04-validation → sniper agent
8. 09-create-pr → gh pr create
1. 00-init-branch → git checkout -b hotfix/xxx
2. 03-execution → Minimal fix only
3. 03.7-verification → Verify fix resolves the issue
4. 04-validation → sniper agent
5. 08-check-test → Run tests
6. 09-create-pr → Urgent merge
🚨 STOP at 90 lines → Split immediately
❌ NEVER exceed 100 lines
📊 Target: 50-80 lines per file
✅ src/interfaces/ (global)
✅ src/types/ (type definitions)
✅ Contracts/ (PHP/Laravel)
❌ NEVER in component files
01-analyze: explore-codebase + research-expert (PARALLEL)
04-validate: sniper (MANDATORY after ANY change)
ALWAYS run 2 agents in parallel:
1. explore-codebase
→ Map project structure
→ Find existing patterns
→ Identify change locations
2. research-expert
→ Verify official documentation
→ Confirm API methods
→ Check best practices
ALWAYS use TaskCreate:
1. Break down into tasks
2. Each task <100 lines
3. Plan file splits FIRST
4. Map dependencies (addBlockedBy)
FOLLOW plan strictly with TDD cycle:
1. Create interfaces FIRST
2. Write failing test (RED) → verify it fails
3. Write minimal code (GREEN) → verify it passes
4. Refactor → keep tests green
5. Monitor file sizes (<100 lines)
6. Write JSDoc/comments
7. Atomic commits per task
See tdd skill for detailed RED-GREEN-REFACTOR rules.
BEFORE sniper, verify functional correctness:
1. Re-read original request
2. List all acceptance criteria
3. Verify each with evidence
4. Check for regressions
5. Confirm: "Problem is RESOLVED"
See verification skill for detailed checklist.
ALWAYS run sniper:
6-phase validation:
1. explore-codebase
2. research-expert
3. grep usages
4. run linters
5. apply fixes
6. ZERO errors
feature/ISSUE-123-short-description
fix/ISSUE-456-bug-name
hotfix/ISSUE-789-urgent-fix
refactor/ISSUE-321-cleanup
docs/ISSUE-654-readme
test/ISSUE-987-coverage
✅ Short-lived branches (1-3 days)
✅ Small, focused changes
✅ Sync frequently with main
✅ Squash and merge
<type>(<scope>): <description>
Types: feat, fix, refactor, docs, test, chore
Scope: component/feature name
Description: imperative mood, <50 chars
feat(auth): add JWT authentication
fix(cart): resolve quantity validation
refactor(api): extract fetch utilities
test(auth): add login component tests
□ sniper passes (ZERO errors)
□ All tests pass
□ Build succeeds
□ Self-review complete
□ No console.logs
□ No TODO unaddressed
□ Files <100 lines
□ Interfaces in correct location
□ JSDoc on all exports
□ No any types
□ Error handling complete
feat(auth): add social login with Google
fix(cart): resolve quantity update bug
refactor(api): extract fetch utilities
□ Summary (1-3 bullets)
□ Changes (added/modified/removed)
□ Related issues (Closes #xxx)
□ Test plan (checkboxes)
□ Screenshots (if UI changes)
START
│
▼
┌───────────────┐
│ 00-init-branch│
└───────┬───────┘
│
▼
┌───────────────┐
│ 00.5-brain- │ ← brainstorming skill (NEW)
│ storm │ questions → design → approval
└───────┬───────┘
│
▼
┌───────────────┐
│ 01-analyze │ ← explore + research
└───────┬───────┘
│
▼
┌───────────────┐
│ 02-plan │ ← TaskCreate
└───────┬───────┘
│
▼
┌───────────────┐
│ 03-execute │ ← TDD: test first (RED→GREEN)
└───────┬───────┘
│
▼
┌───────────────┐
│ 03.5-elicit │ ← expert self-review
└───────┬───────┘
│
▼
┌───────────────┐
│ 03.7-verify │ ← verification skill (NEW)
└───────┬───────┘ functional resolution check
│
▼
┌───────────────┐
│ 04-validate │ ← sniper (code quality)
└───────┬───────┘
│
▼
┌───────────────┐
│ 05-review │
└───────┬───────┘
│
▼
┌───────────────┐
│ 09-create-pr │
└───────┬───────┘
│
▼
DONE
❌ Skip explore-codebase or research-expert
❌ Assume API syntax without verification
❌ Create files >100 lines
❌ Put interfaces in component files
❌ Skip sniper after changes
❌ Merge without tests
❌ Large PRs (>400 lines)
All detailed guides in references/ directory:
references/
├── 00-init-branch.md # Branch creation
├── 01-analyze-code.md # Code analysis
├── 02-features-plan.md # Planning
├── 03-execution.md # Implementation
├── 04-validation.md # Validation
├── 05-review.md # Self-review
├── 06-fix-issue.md # Issue fixes
├── 07-add-test.md # Test writing
├── 08-check-test.md # Test running
└── 09-create-pr.md # PR creation
Framework-specific APEX methodology guides:
| Framework | Directory | Tools |
|---|---|---|
| Laravel | references/laravel/ | Pest, Larastan, Pint |
| Next.js | references/nextjs/ | Vitest, Playwright, ESLint |
| React | references/react/ | Vitest, Testing Library, Biome |
| Swift | references/swift/ | XCTest, SwiftLint, swift-format |
Project Type → References Used
─────────────────────────────────────
composer.json → references/laravel/
next.config.* → references/nextjs/
vite.config.* → references/react/
Package.swift → references/swift/
Default → references/ (generic)
references/[framework]/
├── 00-init-branch.md # Framework-specific branching
├── 01-analyze-code.md # Framework exploration tools
├── 02-features-plan.md # Planning patterns
├── 03-execution.md # SOLID implementation
├── 04-validation.md # Linters and formatters
├── 05-review.md # Framework checklist
├── 06-fix-issue.md # Common fixes
├── 07-add-test.md # Testing patterns
├── 08-check-test.md # Test commands
└── 09-create-pr.md # PR template