Where Are the Key Skill Files Located in the Anthropics knowledge-work-plugins Repository

Key skill files in the anthropics/knowledge-work-plugins repository are located at /<domain>/skills/<skill-name>/SKILL.md, where each domain folder contains a skills/ subdirectory housing individual capability definitions.

The anthropics/knowledge-work-plugins repository organizes every capability as a self-contained skill definition. These key skill files serve as the canonical source for input/output schemas, examples, and auxiliary documentation that the cowork-plugin runtime consumes.

Directory Structure Convention

The repository enforces a strict three-level hierarchy to isolate business domains from implementation details.

Domain Folders

Top-level directories group related business functions. Available domains include product-management, human-resources, engineering, finance, data, customer-support, marketing, legal, partner-built, and bio-research.

Skill Subdirectories

Each domain contains a skills/ folder that aggregates related capabilities. For example, product-management/skills/ contains write-spec, sprint-planning, and roadmap-update.

SKILL.md Files

Every individual skill resides in its own subfolder containing a SKILL.md file. This markdown file is the definitive specification, defining the skill's purpose, input JSON schema, output JSON schema, and usage examples.

Representative Key Skill File Paths

The following table maps domains to their canonical skill definitions:

Domain	Skill	File Path
Product Management	Write-Spec	product-management/skills/write-spec/SKILL.md
Product Management	Sprint-Planning	product-management/skills/sprint-planning/SKILL.md
Human Resources	Recruiting-Pipeline	human-resources/skills/recruiting-pipeline/SKILL.md
Engineering	Incident-Response	engineering/skills/incident-response/SKILL.md
Engineering	Debug	engineering/skills/debug/SKILL.md
Finance	Variance-Analysis	finance/skills/variance-analysis/SKILL.md
Data	SQL-Queries	data/skills/sql-queries/SKILL.md
Partner-Built (Zoom)	Zoom-MCP Whiteboard	partner-built/zoom-plugin/skills/zoom-mcp/whiteboard/SKILL.md

Programmatically Accessing Skill Definitions

Integrations typically load skills by fetching the raw SKILL.md content and parsing its YAML front matter. The following Python example retrieves the Write-Spec skill definition directly from the repository:

import requests
import yaml

# Construct raw GitHub URL for the skill file

SKILL_URL = (
    "https://raw.githubusercontent.com/anthropics/knowledge-work-plugins/main/"
    "product-management/skills/write-spec/SKILL.md"
)

resp = requests.get(SKILL_URL)
resp.raise_for_status()
content = resp.text

# Split YAML front matter from markdown body

front_matter, md_body = content.split("---", 2)[1:3]
spec = yaml.safe_load(front_matter)

print("Skill name:", spec["name"])
print("Description:", spec["description"])
print("Input schema:", spec["input"]["properties"])
print("Output schema:", spec["output"]["properties"])

This script outputs the skill's meta-information, which the cowork-plugin runtime uses to validate prompts and construct execution payloads.

Summary

• Key skill files follow the rigid path convention /<domain>/skills/<skill-name>/SKILL.md.
• Each SKILL.md file contains YAML front matter specifying input/output schemas and a markdown body with implementation guidance.
• Domain folders like engineering/ and product-management/ organize capabilities by business function.
• The repository exposes these definitions via standard GitHub raw content URLs for programmatic consumption.

Frequently Asked Questions

What is the exact file path pattern for skill definitions?

Skill definitions reside at /<domain>/skills/<skill-name>/SKILL.md. The domain folder represents the business vertical, the skills/ subdirectory aggregates capabilities, and the individual skill folder houses the canonical SKILL.md specification file.

How do I add a new skill to the repository?

Create a new subfolder under the appropriate domain's skills/ directory and include a SKILL.md file with valid YAML front matter defining the name, description, input, and output schemas. The cowork-plugin runtime automatically discovers files matching this pattern.

What information does a SKILL.md file contain?

Each SKILL.md contains YAML front matter specifying the skill's name, description, version, and JSON schemas for inputs and outputs, followed by a markdown body with usage examples, constraints, and implementation notes.

Can external tools reference these skill files directly?

Yes, tools can fetch raw content directly from GitHub using URLs like https://raw.githubusercontent.com/anthropics/knowledge-work-plugins/main/<path>/SKILL.md and parse the YAML front matter to discover schema definitions without cloning the repository.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}