split-skill

Extract sub-capabilities from oversized SKILL.md files into child skills.

3|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/dvy1987/agent-loom --skill split-skill
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: split-skill
Source: https://github.com/dvy1987/agent-loom/tree/main/.agents/skills/split-skill
Command: npx skills add https://github.com/dvy1987/agent-loom --skill split-skill

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill reduces oversized SKILL.md files by extracting coherent sub-capabilities into child skills or linking to existing skills, preserving full functionality while keeping each skill under the 200-line guideline. It prevents unnecessary duplication, enforces security scans before splitting, and codifies safe, repeatable split and delegation workflows.

Core Features & Use Cases

  • Discover-and-Link First: Scans the skill library to find existing skills that can be linked or marginally adapted instead of creating new children.
  • Type A & B Extraction: Supports both phase/feature extraction (Type A) and shared-capability extraction across multiple parents (Type B), with clear contracts and AGENTS.md updates.
  • Validation and Safety: Invokes secure-* scanners, verifies line counts, runs agentskills validate, and generates precise commit metadata to ensure regressions are avoided.
  • Use Case: While maintaining a large improve-skills file, use this Skill to extract a repeated "link-check" flow into a shared child skill and update all callers automatically.

Quick Start

Analyze the oversized SKILL.md at .agents/skills/<parent> and produce a recommended action plan that links to existing skills when possible or extracts a new child with a suggested name, inputs, outputs, and commit message.

Frequently Asked Questions about split-skill

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

FAQPage Schema
How do I refactor an oversized SKILL.md file that exceeds 200 lines?

Refactor an oversized SKILL.md file by extracting coherent sub-capabilities into child skills or linking to existing skills, ensuring each file stays under 200 lines while preserving full functionality.

When should I split a monolithic skill into smaller reusable parts?

Split a monolithic skill when it exceeds 200 lines, is duplicated across multiple skills, or contains an identifiable sub-workflow suitable for extraction into a shared child skill.

How do I extract a shared capability that is duplicated across multiple agent skills?

Extract shared capabilities across multiple parents using Type B extraction, which creates a shared child skill with clear contracts and automatically updates all callers in the AGENTS.md call-graph.

What's the best way to find existing skills to link to before creating new child skills?

Use a discover-and-link approach that scans your skill library to find existing skills that can be linked or marginally adapted, preventing unnecessary duplication before extracting new children.

Does splitting a skill require running validation and security scans beforehand?

Yes, splitting requires pre-split security scans using secure-* scanners, verifying line counts, running agentskills validate, and generating precise commit metadata to avoid regressions.