| name | rad-create-skill |
| description | Create new Agent Skills for GitHub Copilot from prompts or by duplicating this template. Use when asked to "create a skill", "make a new skill", "scaffold a skill", or when building specialized AI capabilities with bundled resources. Generates SKILL.md files with proper frontmatter, directory structure, and optional scripts/references/assets folders. |
Make Skill Template
A meta-skill for creating new Agent Skills. Use this skill when you need to scaffold a new skill folder, generate a SKILL.md file, or help users understand the Agent Skills specification.
When to Use This Skill
- User asks to "create a skill", "make a new skill", or "scaffold a skill"
- User wants to add a specialized capability to their GitHub Copilot setup
- User needs help structuring a skill with bundled resources
- User wants to duplicate this template as a starting point
Prerequisites
- Understanding of what the skill should accomplish
- A clear, keyword-rich description of capabilities and triggers
- Knowledge of any bundled resources needed (scripts, references, assets, templates)
Creating a New Skill
Step 1: Create the Skill Directory
Create a new folder with a lowercase, hyphenated name:
skills/<skill-name>/
└── SKILL.md # Required
Step 2: Generate SKILL.md with Frontmatter
Every skill requires YAML frontmatter with name and description:
---
name: <skill-name>
description: '<What it does>. Use when <specific triggers, scenarios, keywords users might say>.'
---
Frontmatter Field Requirements
| Field | Required | Constraints |
|---|
name | Yes | 1-64 chars, lowercase letters/numbers/hyphens only, must match folder name |
description | Yes | 1-1024 chars, single-quoted single-line string — no block scalars (>-, ` |
license | No | License name or reference to bundled LICENSE.txt |
compatibility | No | 1-500 chars, environment requirements if needed |
metadata | No | Key-value pairs for additional properties |
allowed-tools | No | Space-delimited list of pre-approved tools (experimental) |
Reserved namespace. The rad- prefix is reserved for skills shipped by the orchestration system. Do not author downstream skills under the rad-* name; the planner-spawn manifest filter excludes them by design. See CLAUDE.md for the full reservation note.
Description Best Practices
CRITICAL: The description is the PRIMARY mechanism for automatic skill discovery. Include:
- WHAT the skill does (capabilities)
- WHEN to use it (triggers, scenarios, file types)
- Keywords users might mention in prompts
Good example:
description: 'Toolkit for testing local web applications using Playwright. Use when asked to verify frontend functionality, debug UI behavior, capture browser screenshots, or view browser console logs. Supports Chrome, Firefox, and WebKit.'
Poor example:
description: 'Web testing helpers'
Step 3: Write the Skill Body
After the frontmatter, add markdown instructions. Recommended sections:
| Section | Purpose |
|---|
# Title | Brief overview |
## When to Use This Skill | Reinforces description triggers |
## Prerequisites | Required tools, dependencies |
## Step-by-Step Workflows | Numbered steps for tasks |
## Troubleshooting | Common issues and solutions |
## References | Links to bundled docs |
Step 4: Add Optional Directories (If Needed)
| Folder | Purpose | When to Use |
|---|
scripts/ | Executable code (Python, Bash, JS) | Automation that performs operations |
references/ | Documentation agent reads | API references, schemas, guides |
assets/ | Static files used AS-IS | Images, fonts, templates |
templates/ | Starter code agent modifies | Scaffolds to extend |
Step 5: Remove References from Existing Documentation (CRITICAL)
When creating a skill from existing documentation, you MUST systematically remove all references to that documentation from the rest of the codebase. Skills are auto-discovered via frontmatter - duplication wastes context.
Why This Matters
- Context Efficiency: Skills load ONLY when needed via auto-discovery. Documentation references pre-load content unnecessarily.
- Single Source of Truth: All instructions for a capability should live in the skill, not scattered across docs.
- Cleaner Agent Context: Agents work more efficiently with lean, focused documentation.
Systematic Cleanup Process
1. Search for ALL References
Use workspace-wide search (NOT directory-limited) for:
- Original documentation filename (e.g.,
API_CLIENT_GENERATION.md)
- Topic keywords (e.g., "API client generation", "regenerate TypeScript client")
- Links to original docs or skill paths
- Instructional phrases (e.g., "run
pnpm generate:api-client")
2. Remove from Core Documentation
Clean these files completely:
docs/*.md - All main documentation filesCLAUDE.md, AGENTS.md - Agent instruction files.github/copilot-instructions.md - Copilot configuration.github/chatmodes/*.chatmode.md - All chat mode files
3. What to Remove
✅ Remove entirely:
- Instructional content (how-to steps, commands)
- Links to original documentation or skill paths
- Detailed explanations covered by the skill
- Duplicate examples/workflows
✅ Keep (these are OK):
- Brief architectural mentions (e.g., "Generated API client exists")
- Configuration references (e.g.,
package.json scripts needed for execution)
- Historical project documentation (
docs/projects/, docs/features/)
4. Create Redirect File (Optional)
If the original documentation file had many external links, consider leaving a redirect:
# [Original Topic Name]
**This documentation has moved to a specialized skill.**
See: `.claude/skills/<skill-name>/SKILL.md`
This skill is auto-discovered by agents when needed.
5. Use Subagents for Large Cleanups
For comprehensive cleanup:
- Recon subagent: Catalog ALL references (files, lines, context)
- Cleanup subagents: Parallel cleanup of different file categories
- Verification subagent: Final sweep to confirm completeness
Example Cleanup Checklist:
Lessons from API Client Generation Skill Migration
What worked well:
- Using multiple subagents for parallel cleanup
- Creating comprehensive inventory before making changes
- Preserving historical project documentation for archival integrity
- Creating redirect file for backward compatibility
What to avoid:
- Directory-limited searches (use workspace-wide instead)
- Updating links instead of removing (links still waste context)
- Leaving "see skill for details" references (redundant - auto-discovery handles it)
- Touching historical project docs (breaks archival integrity)
Example: Complete Skill Structure
my-awesome-skill/
├── SKILL.md # Required instructions
├── LICENSE.txt # Optional license file
├── scripts/
│ └── helper.py # Executable automation
├── references/
│ ├── api-reference.md # Detailed docs
│ └── examples.md # Usage examples
├── assets/
│ └── diagram.png # Static resources
└── templates/
└── starter.ts # Code scaffold
⚠️ Do NOT Wrap SKILL.md in Code Fences
The examples in this file show skill content inside ```skill code fences for documentation display purposes only. Actual SKILL.md files must be plain Markdown — do not wrap the file content in any code fence. The frontmatter must start at line 1 with ---.
Quick Start: Duplicate This Template
- Copy the
make-skill-template/ folder
- Rename to your skill name (lowercase, hyphens)
- Update
SKILL.md:
- Change
name: to match folder name
- Write a keyword-rich
description:
- Replace body content with your instructions
- Add bundled resources as needed
- Validate with
npm run skill:validate
Validation Checklist
Troubleshooting
| Issue | Solution |
|---|
| Skill not discovered | Improve description with more keywords and triggers |
| Validation fails on name | Ensure lowercase, no consecutive hyphens, matches folder |
| Description too short | Add capabilities, triggers, and keywords |
| Assets not found | Use relative paths from skill root |
References
