Agent skill
ai-pair-programming
Orchestrates a dual-AI engineering loop where a Primary AI plans and implements, while a Review AI validates and reviews, with continuous feedback for optimal code quality. Supports custom AI pairing (Claude, Codex, Gemini, etc.)
Install this agent skill to your Project
npx add-skill https://github.com/Nam0101/my-claude-agent-skill/tree/main/ai-pair-programming
SKILL.md
AI Pair Programming Skill
Overview
A flexible pair programming workflow that allows you to combine any two AI assistants for collaborative development. Configure your preferred AI pair based on their strengths.
AI Configuration
Before using this skill, configure your AI pair by setting the following roles:
| Role | Description | Examples |
|---|---|---|
| Primary AI | Handles architecture, planning, and code implementation | Claude, Gemini, GPT-4 |
| Review AI | Validates plans and reviews code for quality assurance | Codex, Claude, Gemini |
Example Configurations
| Configuration | Primary AI | Review AI | Use Case |
|---|---|---|---|
| Claude + Codex | Claude | Codex | Strong planning + Fast validation |
| Claude + Gemini | Claude | Gemini | Multi-model perspective |
| Gemini + Claude | Gemini | Claude | Leverage Gemini's context + Claude's review |
| GPT-4 + Codex | GPT-4 | Codex | GPT planning + Codex validation |
Setting Up Your AI Pair
Replace {{PRIMARY_AI}} and {{REVIEW_AI}} placeholders throughout this skill with your chosen AI names.
Example:
{{PRIMARY_AI}}→Claude{{REVIEW_AI}}→Codex
For command-line tools, replace {{REVIEW_AI_CLI}} with the appropriate CLI command:
- Codex:
codex exec --sandbox read-only - Claude:
claude --print - Gemini:
gemini - Other: Use your AI's CLI tool
Note: CLI commands are examples. Please verify the correct CLI syntax for your specific AI tools and their current versions.
Core Workflow Philosophy
This skill implements a balanced engineering loop:
- {{PRIMARY_AI}}: Architecture, planning, and execution
- {{REVIEW_AI}}: Validation and code review
- Continuous Review: Each AI reviews the other's work
- Context Handoff: Always continue with whoever last cleaned up
Note: Model + reasoning settings are assumed to be handled by each AI's existing config. Commands below use generic placeholders.
Phase 1: Planning with {{PRIMARY_AI}}
- Start by creating a detailed plan for the task
- Break down the implementation into clear steps
- Document assumptions and potential issues
- Output the plan in a structured format
Phase 2: Plan Validation with {{REVIEW_AI}}
- Send the plan to {{REVIEW_AI}} for validation:
echo "Review this implementation plan and identify any issues:
[{{PRIMARY_AI}}'s plan here]
Check for:
- Logic errors
- Missing edge cases
- Architecture flaws
- Security concerns" | {{REVIEW_AI_CLI}}
- Capture {{REVIEW_AI}}'s feedback
Phase 3: Feedback Loop
If {{REVIEW_AI}} finds issues:
- Summarize {{REVIEW_AI}}'s concerns to the user
- Refine the plan based on feedback
- Ask user (via
AskUserQuestion): "Should I revise the plan and re-validate, or proceed with fixes?" - Repeat Phase 2 if needed
Phase 4: Execution
Once the plan is validated:
- {{PRIMARY_AI}} implements the code using available tools (Edit, Write, Read, etc.)
- Break down implementation into manageable steps
- Execute each step carefully with proper error handling
- Document what was implemented
Phase 5: Cross-Review After Changes
After every change:
- Send {{PRIMARY_AI}}'s implementation to {{REVIEW_AI}} for review:
- Bug detection
- Performance issues
- Best practices validation
- Security vulnerabilities
- {{PRIMARY_AI}} analyzes {{REVIEW_AI}}'s feedback and decides:
- Apply fixes immediately if issues are critical
- Discuss with user if architectural changes needed
- Document decisions made
Phase 6: Iterative Improvement
- After {{REVIEW_AI}} review, {{PRIMARY_AI}} applies necessary fixes
- For significant changes, send back to {{REVIEW_AI}} for re-validation
- Continue the loop until code quality standards are met
- For AI tools that support session continuation, use the resume flag:
echo "Review the updated implementation" | {{REVIEW_AI_CLI}} --resume
Note: Session resumption support varies by AI tool. For tools without resume support, start a new session with full context.
Recovery When Issues Are Found
When {{REVIEW_AI}} identifies problems:
- {{PRIMARY_AI}} analyzes the root cause
- Implements fixes using available tools
- Sends updated code back to {{REVIEW_AI}} for verification
- Repeats until validation passes
When implementation errors occur:
- {{PRIMARY_AI}} reviews the error/issue
- Adjusts implementation strategy
- Re-validates with {{REVIEW_AI}} before proceeding
Best Practices
- Always validate plans before execution
- Never skip cross-review after changes
- Maintain clear handoff between AIs
- Document who did what for context
- Use resume to preserve session state (check your AI tool's documentation for support)
Command Reference
| Phase | Command Pattern | Purpose |
|---|---|---|
| Validate plan | echo "plan" | {{REVIEW_AI_CLI}} |
Check logic before coding |
| Implement | {{PRIMARY_AI}} uses Edit/Write/Read tools | {{PRIMARY_AI}} implements the validated plan |
| Review code | echo "review changes" | {{REVIEW_AI_CLI}} |
{{REVIEW_AI}} validates {{PRIMARY_AI}}'s implementation |
| Continue review | echo "next step" | {{REVIEW_AI_CLI}} --resume |
Continue validation session (if supported) |
| Apply fixes | {{PRIMARY_AI}} uses Edit/Write tools | {{PRIMARY_AI}} fixes issues found by {{REVIEW_AI}} |
| Re-validate | echo "verify fixes" | {{REVIEW_AI_CLI}} |
{{REVIEW_AI}} re-checks after fixes |
Note: The
--resumeflag is supported by some AI tools (e.g., Codex). For tools without resume support, start a new session with the necessary context.
Error Handling
- Stop on non-zero exit codes from {{REVIEW_AI}}
- Summarize {{REVIEW_AI}} feedback and ask for direction via
AskUserQuestion - Before implementing changes, confirm approach with user if:
- Significant architectural changes needed
- Multiple files will be affected
- Breaking changes are required
- When {{REVIEW_AI}} warnings appear, {{PRIMARY_AI}} evaluates severity and decides next steps
The Perfect Loop
Plan ({{PRIMARY_AI}}) → Validate Plan ({{REVIEW_AI}}) → Feedback →
Implement ({{PRIMARY_AI}}) → Review Code ({{REVIEW_AI}}) →
Fix Issues ({{PRIMARY_AI}}) → Re-validate ({{REVIEW_AI}}) → Repeat until perfect
This creates a self-correcting, high-quality engineering system where:
- {{PRIMARY_AI}} handles all code implementation and modifications
- {{REVIEW_AI}} provides validation, review, and quality assurance
Quick Start Examples
Example 1: Claude + Codex (Original Configuration)
PRIMARY_AI = Claude
REVIEW_AI = Codex
REVIEW_AI_CLI = codex exec --sandbox read-only
Example 2: Claude + Gemini
PRIMARY_AI = Claude
REVIEW_AI = Gemini
REVIEW_AI_CLI = gemini
Example 3: Gemini + Claude
PRIMARY_AI = Gemini
REVIEW_AI = Claude
REVIEW_AI_CLI = claude --print
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
figma-to-compose
Generate Android Jetpack Compose UI from Figma using Figma Desktop MCP (get_metadata, get_variable_defs, create_design_system_rules, get_design_context, get_screenshot). Automatically detect icon/vector nodes in Figma, obtain SVG/path data when available, and convert icons to Android VectorDrawable XML using Android MCP Toolkit (convert-svg-to-android-drawable). Use when the user shares a Figma link/node-id and asks to implement UI in Compose from Figma.
git-guardrails-claude-code
Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.
setup-pre-commit
Set up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests in the current repo. Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing.
handoff
Compact the current conversation into a handoff document for another agent to pick up.
edit-article
Edit and improve articles by restructuring sections, improving clarity, and tightening prose. Use when user wants to edit, revise, or improve an article draft.
obsidian-vault
Search, create, and manage notes in the Obsidian vault with wikilinks and index notes. Use when user wants to find, create, or organize notes in Obsidian.
Didn't find tool you were looking for?