fanout

Splits files across parallel agents to apply one mechanical instruction tree-wide.

Updated Jun 28, 2026
One-click install
npx skills add https://github.com/Cohey0727/CodingAgentTools --skill fanout-cohey0727
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fanout
Source: https://github.com/Cohey0727/CodingAgentTools/tree/main/skills/fanout
Command: npx skills add https://github.com/Cohey0727/CodingAgentTools --skill fanout-cohey0727

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Applying the same mechanical change or scan across dozens or hundreds of files is slow and error-prone when done sequentially by a single agent. This Skill divides a file set into disjoint slices, dispatches parallel worker agents in one message, and aggregates their reports. ## Core Features & Use Cases - Parallel file slicing: Enumerates tracked files with git ls-files, sorts by size, and round-robins them into up to 6 worker buckets so no file is touched twice. - Detect, fix, and dry-run modes: Run read-only scans, apply edits, or preview proposed changes before committing to a fix. - Structured aggregation: Each worker returns a changed/skipped/no-op report that the orchestrator merges into a single Markdown summary with cross-cutting notes. - Use Case: Replace inline StyleSheet.create blocks across 80 React Native components by dispatching 4 workers, each editing only its assigned slice, then reviewing the merged report before running CI. ## Quick Start Ask the agent to run /fanout with an instruction like "find raw Pressable used as buttons" over your source directory in detect mode.

Frequently Asked Questions about fanout

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I apply the same code change across many files in parallel?▼

Invoke /fanout with a concrete instruction, a root directory, and an optional glob filter. The skill enumerates tracked files, splits them into disjoint slices, and dispatches up to 6 parallel agents that each edit only their assigned files.

How do I scan a codebase for a pattern grep cannot find?▼

Run /fanout in detect mode with a natural-language description of the pattern, such as semantic misuse of a component. Workers read their file slices and report findings as path:line entries without making any edits.

When should I not use parallel fan-out for a refactor?▼

Avoid fan-out for cross-file refactors that need shared state, such as renaming a function and updating all callers. Those tasks belong to a single agent; fan-out only works when each file can be handled independently.

Can I preview changes before workers edit files?▼

Yes, pass --dry-run so each worker describes the change it would make per file without editing. You can also use detect mode for a read-only report before committing to a fix run.

What is the maximum number of parallel workers?▼

The worker count is hard-capped at 6, chosen automatically from file count when not specified. More workers increase context cost and parallel-tool strain without improving wall-clock time.