Claude Code Plugins

Community-maintained marketplace

Feedback

pair

@ratacat/pair-programmer-ai
0
0

Launch a pair programming agent to watch your work and catch issues

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name pair
description Launch a pair programming agent to watch your work and catch issues
triggers /pair, pair program with me, start pair programming, launch pair programmer

Pair Programming Skill

Launch a pair programming session where a second AI agent watches your work and provides real-time feedback on bugs, security issues, and missed edge cases.

Commands

  • /pair or /pair start - Start a pair programming session
  • /pair stop - End the current session
  • /pair status - Check if pairing is active
  • /pair --backend <name> - Start with specific backend (claude-opus or codex)

How It Works

  1. Bridge Server: A Unix socket server routes messages between you (main agent) and the pair agent
  2. Activity Streaming: Your tool calls are streamed to the pair agent in real-time
  3. Feedback Injection: Pair agent feedback appears before your next action

Starting a Session

When the user invokes /pair or asks to pair program:

  1. Generate a unique session ID
  2. Start the bridge server:
    pair-bridge start
  3. Spawn the pair agent using the Task tool with run_in_background: true
  4. Confirm to user that pairing is active

Example spawn prompt for the pair agent:

You are a pair programmer watching another agent work. Connect to the bridge
at /tmp/claude-pair-{session_id}.sock and run the watch loop:

1. Use `pair-bridge wait` to block until activity
2. Analyze each tool call for issues
3. Use `pair-bridge emit feedback '{"severity":"high|medium|low","message":"..."}' when you spot problems
4. Stay quiet on routine operations

Focus on: bugs, security issues, missed edge cases, API contract changes.

Stopping a Session

When the user says /pair stop or ends pairing:

  1. Stop the bridge server:
    pair-bridge stop
  2. The pair agent will receive a stop signal and exit
  3. Summarize what the pair observed (activity count, feedback given)

Backend Selection

The pair agent can be:

  • claude-opus (default): Spawns Opus via Task tool
  • codex: Spawns OpenAI Codex via CLI

Set default in config or override with --backend:

/pair --backend codex

Implementation Notes

  • Hooks (pair-emit-activity.sh, pair-check-feedback.sh) handle the main agent side
  • The skill's job is session lifecycle, not message passing
  • Pair agent runs in background and may outlive individual tool calls
  • Session state is in the bridge server, not the skill

Example Session

User: /pair

Agent: Starting pair programming session...
  - Bridge started on /tmp/claude-pair-abc123.sock
  - Launching pair programmer (claude-opus)
  - Pair programming active!

Your pair will watch your work and provide feedback on issues.

User: Refactor the auth module

Agent: [Works on refactoring...]

<pair-feedback>
PAIR PROGRAMMER FEEDBACK:
You removed the rate limiting middleware - this was protecting against brute force attacks.
Location: src/auth/login.ts:45
SEVERITY: HIGH - Address before continuing
</pair-feedback>

Agent: Good catch! Let me restore that rate limiting...

User: /pair stop

Agent: Ending pair programming session...
  - Session lasted 12 minutes
  - 34 tool calls observed
  - 2 feedback items provided (1 high, 1 medium)
  - All high-priority items addressed