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.)

name: ai-pair-programming description: 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.)

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.

Before using this skill, configure your AI pair by setting the following roles:

Replace{{PRIMARY_AI}}and{{REVIEW_AI}}placeholders throughout this skill with your chosen AI names.

- {{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.

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.
- 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}}

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}}

- Summarize {{REVIEW_AI}}'s concerns to the user
- Refine the plan based on feedback
- Ask user (viaAskUserQuestion): "Should I revise the plan and re-validate, or proceed with fixes?"
- Repeat Phase 2 if needed
- {{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
- Send {{PRIMARY_AI}}'s implementation to {{REVIEW_AI}} for review:

- Bug detection
- Performance issues
- Best practices validation
- Security vulnerabilities

- Apply fixes immediately if issues are critical
- Discuss with user if architectural changes needed
- Document decisions made
- 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.
- {{PRIMARY_AI}} analyzes the root cause
- Implements fixes using available tools
- Sends updated code back to {{REVIEW_AI}} for verification
- Repeats until validation passes
- {{PRIMARY_AI}} reviews the error/issue
- Adjusts implementation strategy
- Re-validates with {{REVIEW_AI}} before proceeding

- Always validate plansbefore execution
- Never skip cross-reviewafter changes
- Maintain clear handoffbetween AIs
- Document who did whatfor context
- Use resumeto preserve session state (check your AI tool's documentation for support)

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.
- Stop on non-zero exit codes from {{REVIEW_AI}}
- Summarize {{REVIEW_AI}} feedback and ask for direction viaAskUserQuestion
- Before implementing changes, confirm approach with user if:

- Significant architectural changes needed
- Multiple files will be affected
- Breaking changes are required

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

Example 1: Claude + Codex (Original Configuration)

PRIMARY_AI = Claude REVIEW_AI = Codex REVIEW_AI_CLI = codex exec --sandbox read-only

PRIMARY_AI = Claude REVIEW_AI = Gemini REVIEW_AI_CLI = gemini

PRIMARY_AI = Gemini REVIEW_AI = Claude REVIEW_AI_CLI = claude --print

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

AI-to-AI code review platform — Claude, Codex, and Gemini cross-check each other via MCP, REST API, and CLI for consensus-based results.

Provides AI assistants with a comprehensive, one-time analysis for complete codebase context and understanding.

Automate Google Jules, the AI coding assistant, for tasks like code reviews, repository management, and AI-powered development workflows.

Analyze and visualize code graphs using CodeGPT.

Hybrid decomp tool — AI decides WHERE to split your monolith, deterministic engine COPIES the code. Minimize tokens, maximize syntax validity.

An intelligent security agent that analyzes AI-generated code for vulnerabilities and assists with remediation.

ast-mcp-server gives coding agents compact, type-aware access to TypeScript and JavaScript projects. It uses the real compiler project model through ts-morph, so declarations, references, rename locations, and diagnostics come from the AST instead of text-search guesses.

An MCP server to help AI assistants to answer questions and generate AccelByte Extend SDK code more effectively .

Instead of direct calling MCP tools, mcpcode server transforms MCP tool calls into TypeScript programs, enabling smarter, lower-latency orchestration by LLMs.

Skene is a codebase analysis toolkit for product-led growth. It scan your codebase, detect growth opportunities, and generate actionable implementation plans.