Agent skill
XP
Extreme Programming workflow orchestrator. Use when implementing features, adding functionality, or doing test-driven development. Coordinates planning, TDD, refactoring, and commits.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/development/xp
SKILL.md
Extreme Programming Workflow
Overview
This skill orchestrates the full XP workflow for feature implementation. It uses reference documentation for interactive phases and delegates to agents for autonomous tasks.
Getting Started
- Detect project type from files in working directory
- Read the language skill for the detected type:
build.sbtor*.scala→scala-developerbuild.gradle.ktsor*.kt→kotlin-developerGemfileor*.rb→ruby-developer*.uor.unison/→unison-development
- Check for project CLAUDE.md — may contain project-specific guidance that supplements or overrides language defaults
- Begin with PLAN phase
The XP Workflow
┌─────────────────────────────────────────────────────────────┐
│ 📋 PLAN → Discuss and break down the feature │
│ 🔴 DEVELOP → TDD cycle (red-green) │
│ 🔵 REFACTOR → Improve design (tests stay green) │
│ 🔍 REVIEW → Autonomous code review (optional) │
│ 💾 COMMIT → Save working state │
│ 🔁 ITERATE → Next task or feature complete │
└─────────────────────────────────────────────────────────────┘
Phase 1: Planning (📋 PLAN)
Goal: Understand and decompose the feature before writing any code.
Type: Interactive (requires user discussion)
Reference: See references/planning.md
Supporting Skills
glossary— Define unfamiliar domain terms encountered during discussion
Checkpoint
- Requirements understood
- Domain terms added to glossary
- Tasks broken into vertical slices
- First task agreed with user
Phase 2: Development (🔴 DEVELOP)
Goal: Implement the task using strict TDD.
Type: Interactive (user sees tests and implementation)
Reference: See references/development.md + language skill
The TDD Cycle
🔴 RED → Write ONE failing test
🟢 GREEN → Write MINIMUM code to pass
✅ VERIFY → Run all tests, confirm green
Checkpoint
- Test written and failing
- Minimum code makes test pass
- All tests green
Phase 3: Refactoring (🔵 REFACTOR)
Goal: Improve code design while keeping tests green.
Type: Interactive (user approves changes)
Reference: See references/refactor.md
Checkpoint
- All tests pass before refactoring
- One transformation at a time
- All tests pass after each change
- STOP — Ask user if they want further changes
Phase 4: Review (🔍 REVIEW) — Optional
Goal: Autonomous quality check before committing.
Type: Autonomous (runs in isolation)
Invoke: Delegate to code-reviewer agent
When to Use
- Before merging feature branches
- After significant refactoring
- When requested by user
- For complex or security-sensitive changes
Supporting Skills
bugmagnet— Deep test coverage analysis if review identifies gaps
Agent Properties
- Model: Opus (thorough analysis)
- Tools: Read-only (cannot modify code)
- Output: Structured findings report
Checkpoint
- Review any 🔴 CRITICAL findings
- Address 🟡 WARNINGs if time permits
- Note ℹ️ SUGGESTIONs for future
- Run
bugmagnetif test gaps identified
Phase 5: Commit (💾 COMMIT)
Goal: Save working state with clear, conventional commit message.
Type: Autonomous (generates message for approval)
Invoke: Delegate to commit-helper agent
When to Commit
- After each passing test (small commits)
- After completing a task
- After refactoring session
- Before switching branches
Agent Properties
- Model: Haiku (fast and cheap)
- Tools: Bash only (git commands)
- Output: Conventional commit message
Checkpoint
- Changes staged
- Review suggested commit message
- Confirm or adjust message
- Commit created
Phase 6: Iterate (🔁 ITERATE)
Goal: Continue until feature complete.
Type: Interactive (planning next steps with user)
- Mark task as done
- Review remaining tasks
- Adjust plan if needed (new learnings)
- Return to Phase 2 for next task
- When all tasks complete → feature done
Supporting Skills
vault— Log significant learnings or decisions to project notes
Checkpoint
- Task marked complete
- Learnings captured (if significant)
- Remaining tasks reviewed
- Next task selected or feature complete
Announcing Phase Transitions
When switching phases, announce clearly:
📋 PLAN → Starting feature discussion
🔴 DEVELOP → Writing failing test for [behaviour]
🟢 DEVELOP → Making test pass
🔵 REFACTOR → Improving [aspect]
🔍 REVIEW → Delegating to code-reviewer agent
💾 COMMIT → Delegating to commit-helper agent
🔁 ITERATE → Moving to next task
✅ COMPLETE → Feature done
Component Summary
| Phase | Component | Type | User Interaction |
|---|---|---|---|
| PLAN | references/planning.md |
Reference | Discussion required |
| PLAN | glossary skill |
Skill | Define terms |
| DEVELOP | references/development.md |
Reference | Sees tests/code |
| DEVELOP | Language skill (detected in Getting Started) | Skill | Build/test commands |
| REFACTOR | references/refactor.md |
Reference | Approves changes |
| REVIEW | code-reviewer agent |
Agent | Reviews report |
| REVIEW | bugmagnet skill |
Skill | Deep coverage (optional) |
| COMMIT | commit-helper agent |
Agent | Confirms message |
| ITERATE | vault skill |
Skill | Capture learnings |
Core Principles (Always Apply)
- Communication first — discuss before coding
- Small steps — one task, one test, one change at a time
- Continuous feedback — tests run constantly
- Simplicity — implement only what's needed now
- Courage — refactor fearlessly (tests protect you)
- Quality gates — review before merge, commit after green
- Knowledge capture — document learnings and domain terms
Didn't find tool you were looking for?