How does this skill manage the file system for parallel tasks?

It uses Git worktrees to create isolated environments for each subagent, preventing file conflicts and allowing concurrent execution on different branches.

When is Parallel Execute automatically triggered?

It is auto-triggered via the /cat:work command whenever a task size estimate exceeds the context threshold (typically 80K tokens), following an initial decomposition step.

What happens if one subagent fails during a parallel run?

The skill follows a 'continue on failure' strategy, collecting results from successful agents while marking dependent tasks as blocked and reporting the failure details for recovery.

How does it handle dependencies between tasks?

The skill performs a dependency analysis to group tasks. Independent tasks are run concurrently, while dependent tasks are queued until their prerequisites are merged.

Parallel Execute

Name: Parallel Execute
Author: cowwoc

bycowwoc

•

Productivity & Workflow

Orchestrates multiple independent AI subagents concurrently to maximize development throughput and bypass context limits.

Parallel Execute is a high-performance orchestration skill designed for the CAT framework that enables Claude to manage multiple subagents simultaneously. It automatically triggers when tasks are decomposed into independent units, handling the entire lifecycle of spawning sub-processes, monitoring progress in real-time, and merging results back into the main branch. This skill is critical for large-scale engineering tasks that exceed context window thresholds, allowing for isolated execution and significant reductions in wall-clock time for complex implementations.

Key Features

0156 GitHub stars

02Dependency-aware result collection and sequential branch merging

03Automated triggering from large-task decomposition workflows

04Concurrent subagent orchestration for independent work streams

05Real-time monitoring with mandatory progress and token usage tracking

06Isolated Git worktree management for parallel subagent execution

Use Cases

01Simultaneously implementing independent architectural components like lexers, parsers, and documentation

02Bypassing LLM context limits by splitting massive tasks into parallelizable sub-units

03Accelerating large-scale refactoring projects by distributing work across multiple AI agents

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cowwoc/cat parallel-execute

For use in Claude.ai and ChatGPT

Download Skill