create-skill
/create-skill
Create, validate, and install custom local skills. Use when you need to create a new skill (or update an existing skill) that extends agent capabilities with specialized knowledge, workflows, or tool integrations. Also use when asked to add a reusable capability that should persist across sessions.
- Author
- aomni-com ↗
- Repository
- open-duet ↗
- Licence
- MIT
- Last repository change
- 30.10.2026
Assessment and security
- Files in this package
- 5
- Scripts
- 2
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 30.10.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
- 2 scripts in this package. Review them before installing, as Claude may execute them.
Scripts: scripts/init_skill.py, scripts/quick_validate.py
What it does and when to use it
Create, validate, and install custom local skills. Use when you need to create a new skill (or update an existing skill) that extends agent capabilities with specialized knowledge, workflows, or tool integrations. Also use when asked to add a reusable capability that should persist across sessions.
Original author description
Create, validate, and install custom local skills. Use when you need to create a new skill (or update an existing skill) that extends agent capabilities with specialized knowledge, workflows, or tool integrations. Also use when asked to add a reusable capability that should persist across sessions.
How it works
- What Skills Provide
- Core Principles
- Skill Anatomy
- Creation Process
- Managing Skills
- What NOT to Include
SKILL.md preview
--- name: create-skill target: both description: Create, validate, and install custom local skills. Use when you need to create a new skill (or update an existing skill) that extends agent capabilities with specialized knowledge, workflows, or tool integrations. Also use when asked to add a reusable capability that should persist across sessions. --- # Creating Skills Skills are modular packages that extend agent capabilities with specialized knowledge, workflows, and tools. They transform a general-purpose agent into a specialist equipped with procedural knowledge no model fully possesses. ## What Skills Provide 1. **Specialized workflows** — multi-step procedures for specific domains 2. **Tool integrations** — instructions for working with APIs or file formats 3. **Domain expertise** — company-specific knowledge, schemas, business logic 4. **Bundled resources** — scripts, references, and assets for repetitive tasks ## Core Principles **Concise is key.** The context window is shared. Only add what the agent doesn't already know. Prefer concise examples over verbose explanations. **Set appropriate degrees of freedom:** - **High freedom** (text instructions): when multiple approaches are valid - **Medium freedom** (pseudocode/scripts with params): when a preferred pattern exists - **Low freedom** (exact scripts): when operations are fragile or consistency is critical **Progressive disclosure.** Skills load in three levels: 1. Metadata (name + description) — always in context (~100 words) 2. SKILL.md body — loaded when skill triggers (<500 lines) 3. Bundled resources — loaded as needed (unlimited) ## Skill Anatomy ``` skill-name/ ├── SKILL.md # Required — entry point with YAML frontmatter ├── scripts/ # Optional — executable code (Python/Bash) ├── references/ # Optional — docs loaded into context as needed └── assets/ # Optional — templates, images, fonts for output ``` ### SKILL.md - **Frontmatter** (YAML): `name` and `description` are required. Only these two fields determine when the skill triggers — make the description comprehensive about when to use it. - **Body** (Markdown): Instructions and guidance. Only loaded after triggering. ### Bundled Resources - **`scripts/`**: Deterministic code that would otherwise be rewritten each time. Token efficient — can be executed without loading into context. - **`references/`**: Documentation loaded on demand. Keeps SKILL.md lean. For large files (>10k words), include grep patterns in SKILL.md. - **`assets/`**: Files used in output (templates, images, fonts). Not loaded into context. **Avoid duplication** between SKILL.md and reference files. Keep only essential workflow instructions in SKILL.md; move detailed reference material to `references/`. ## Creation Process ### Step 1: Understand the Use Cases Clarify concrete examples of how the skill will be used: - What triggers it? What would a user say? - What functionality should it support? - What's the expected output? ### Step 2: Plan Reusable Contents For each use case, identify what would help when repeating the workflow: - Scripts that would otherwise be rewritten each time - Reference docs (schemas, API docs) that would need rediscovery - Assets (templates, boilerplate) that stay consistent ### Step 3: Initialize A skill you write belongs to you: it lives in your own agent folder, syncs, and travels when you are shared or installed elsewhere. Your `your-folder` system prompt names the path. Scaffold it there: ```bash mkdir -p <your folder>/.agents/skills && cd <your folder>/.agents/skills bunx skills init <skill-name> ``` This creates `<skill-name>/SKILL.md` with a template. Edit it to fill in the content. For a fuller scaffold with example resource directories: ```bash python scripts/init_skill.py <skill-name> --path <your folder>/.agents/skills ``` Write a skill for the whole workspace under the drive root's `.agents/skills` instead — every agent in the workspace loads it — and only when the user asks for that. ### Step 4: Write the Skill **Frontmatter rules:** - `name`: kebab-case, max 64 chars - `description`: max 1024 chars, no angle brackets. Include ALL "when to use" info here — the body only loads after triggering. **Body guidelines:** - Use imperative form - Keep under 500 lines — split to reference files if longer - Include exact commands, code snippets, and examples - Reference bundled files with clear descriptions of when to read them **Design patterns for content organization:** - **Multi-step processes**: See `references/workflows.md` - **Output format/quality standards**: See `references/output-patterns.md` ### Step 5: Validate Run the validation script to check the skill structure: ```bash python scripts/quick_validate.py <your folder>/.agents/skills/<skill-name> ``` Checks: frontmatter format, naming conventions, description length, allowed properties. ### Step 6: Verify It Loads Written in place, the file *is* the installation — there is nothing to register. Confirm the next session will see it: ```bash duet skills ``` Report the skill as ready only once it appears there, with `"scope": "project"` and a path inside your folder. `"scope": "user"` means a machine-level copy, which neither syncs nor travels with you. **Never hand-write into `~/.agents/skills` or `~/.claude/skills`.** Those are managed link farms and edits there are repaired away. A skill that genuinely belongs to the machine rather than to you is installed with `bunx skills add <path> -g -y`, and only when the user asks for that. ### Step 7: Iterate Use the skill on real tasks, notice struggles, update SKILL.md or resources, and test again. ## Managing Skills ```bash duet skills # What this session actually loads, with paths bunx skills remove <name> -y # Remove one from the folder you run it in ``` Deleting the skill's folder removes a hand-written one. ## What NOT to Include Do not create extraneous files: README.md, INSTALLATION_GUIDE.md, CHANGELOG.md, etc. A skill should only contain what the agent needs to do the job.
Files in this package (5)
- SKILL.md 6 KB
- references/output-patterns.md 2 KB
- references/workflows.md 1 KB
- scripts/init_skill.py 11 KB
- scripts/quick_validate.py 3 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · aomni-com/open-duet | 100 | MIT |
| mgiovani/cc-arsenal | 100 | MIT |
| zylos-ai/zylos-core | 100 | MIT |
| mthines/agent-skills | 100 | MIT |
| jongio/skills | 100 | MIT |
| fairladyz625/harness-anything | 100 | AGPL-3.0 |
| coco-research/coco | 100 | MIT |
| builderio/agent-native | 100 | MIT |
| BrandeisPatrick/silo-harness | 100 | MIT |
Database snapshot: 13.11.2026 · package version f9e2c4b7149d (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends the agent's capabilities with specialized knowledge, workflows, or tool integrations. - 90skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Codex's capabilities with specialized knowledge, workflows, or tool integrations. - 100skill-creator
Guide for creating effective skills. Use when creating a new skill (or updating an existing skill) that extends an agent’s capabilities with specialized knowledge, workflows, or tool integrations. - 100skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations. - 100skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations. - 100skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.