forky
Fork Claude sessions to handle side tasks in parallel. Use when the user wants to spawn a parallel Claude session, delegate a task to a background agent, monitor forked sessions, or when they mention "forky", "fork", "spawn", or "parallel task".
When & Why to Use This Skill
Forky is a high-performance parallel session manager designed to enhance productivity by spawning background Claude agents. It solves the bottleneck of sequential task execution by allowing users to delegate side tasks—such as code reviews, research, and documentation—to independent sessions that run in parallel, ensuring a non-blocking and efficient workflow.
Use Cases
- Parallel Code Review: Initiate a background agent to perform a comprehensive review of your code changes while you continue with active development in the main session.
- Concurrent Technical Research: Delegate deep-dive research into libraries or APIs to a forked session, allowing you to focus on implementation while the agent gathers data.
- Automated Test Generation: Spawn a background task to write unit tests or integration tests for newly written code without interrupting your current coding flow.
- Asynchronous Documentation: Offload the creation of technical documentation or README updates to a parallel agent to maintain project velocity.
- Background Monitoring: Use the observability UI and list commands to track multiple long-running tasks across different forked sessions simultaneously.
| name | forky |
|---|---|
| description | Fork Claude sessions to handle side tasks in parallel. Use when the user wants to spawn a parallel Claude session, delegate a task to a background agent, monitor forked sessions, or when they mention "forky", "fork", "spawn", or "parallel task". |
Forky - Parallel Claude Session Manager
Forky lets you spawn parallel Claude sessions to handle side tasks without interrupting your main workflow.
CRITICAL: Never Wait for Forks
Forks are fire-and-forget. After spawning, IMMEDIATELY continue with the next task. NEVER:
- Say "I'll wait for the fork to complete"
- Poll the fork status
- Defer work pending fork completion
Forks notify you when done. Just spawn and move on.
Spawning a Fork
ALWAYS run forky spawn as a background shell operation. This ensures you don't block waiting for the spawn to complete.
# Use run_in_background=true with the Bash tool
~/bin/forky spawn "your task description here"
The binary is at ~/bin/forky (NOT ~/.forky/bin/).
Model Selection
Always use Opus (the default). Do NOT use -m haiku or -m sonnet without asking the user first.
When to Fork
Good:
- Code review while continuing development
- Research while implementing
- Test generation for code you just wrote
- Documentation tasks
- Any independent background work
Avoid:
- Tasks requiring immediate user input
- Modifying same files (use
--worktreeif needed) - Very short tasks (overhead > benefit)
Monitoring
~/bin/forky serve # Start observability UI
~/bin/forky list forks # List all forks
~/bin/forky messages <id> # View fork messages
For Forked Agents
If you ARE a fork (check your callback instruction), call done when finished:
~/bin/forky done <your-fork-id> "summary of what was accomplished"