Dev & Engineering skill-authoringskill-packagingprogressive-disclosureprompt-engineeringpython-scriptsdocumentation

Skill Creator

A meta-skill that guides you through writing, packaging, and iterating high-quality custom Skills for Claude, turning authoring best practices into a repeatable process.

FollowSkills review · FSRS-1.0
Strongly recommended
70/ 100 5-point scale 3.5 / 5
This review was completed under FSRS 1.0. Dimension scores are not converted; it is queued for re-review under FSRS 2.0.
1 2 3 4 5 6
1Utility15 / 20 · 3.8/5

A bundled copy of Anthropic's official 'how to create a skill' guide — pure methodology content (conciseness principle, progressive disclosure), no execution risk.

2Reliability14 / 20 · 3.5/5
3Safety21 / 25 · 4.2/5
4Evidence6 / 15 · 2.0/5
5Usability8 / 10 · 4.0/5
6Maintenance6 / 10 · 3.0/5
Evidence confidence:Low Reviewed Jul 17, 2026
Review evidence [1]
See the full review method →

What does this skill do, and when should you use it?

skill-creator is one of five skills bundled with the Open Cowork open-source desktop app, located at .claude/skills/skill-creator/. It performs no business task itself; instead it teaches Claude how to build new skills: clarifying requirements with concrete examples, planning reusable resources, initializing via init_skill.py, writing SKILL.md under progressive-disclosure principles, and packaging with package_skill.py into a distributable .skill file. It encodes Anthropic-style authoring guidance — concise context, tuned degrees of freedom, and three-level loading. Best suited to advanced users who want to codify team workflows as reusable skills.

When triggered, the skill walks Claude through a six-step process: (1) clarify the skill's purpose and triggers using concrete usage examples; (2) analyze which content belongs in scripts/, references/, or assets/; (3) run scripts/init_skill.py to generate a template with standard directory structure; (4) edit SKILL.md and resources following its conventions — YAML frontmatter limited to name and description, body under 500 lines, progressive disclosure; (5) run scripts/package_skill.py to validate and package into a .skill file (a zip with a .skill extension); (6) iterate based on real usage. It also references two design-pattern guides: references/workflows.md and references/output-patterns.md.

  1. A knowledge manager who wants to wrap company report templates and business rules into a skill so every team member's Claude follows the same output process
  2. A developer building a skill for repetitive coding tasks (scaffolding, fixed-format config files) to avoid re-explaining requirements each time
  3. A user with a messy hand-written skill who wants it restructured to conventions — proper frontmatter, trigger descriptions, and directory layout
  4. A team distributing skills across models/clients, using package_skill.py to produce standardized .skill packages
  5. An Open Cowork desktop user who wants to add a custom skill beyond the built-in pptx/docx/pdf/xlsx set

What are this skill's strengths and limitations?

Pros
  • Complete toolchain from template generation to validated packaging reduces authoring errors
  • Encodes mature best practices: progressive disclosure, context economy, degrees-of-freedom tuning
  • package_skill.py auto-validates frontmatter, naming conventions, and file structure before packaging
  • Six-step process plus reference guides give first-time skill authors a clear path
Limitations
  • The bundled init_skill.py and package_skill.py require a Python runtime; doc-only clients cannot execute them
  • The skill mandates excluding README/CHANGELOG and other auxiliary files, which some teams may find restrictive
  • Whether the .skill output format (a renamed zip) imports into clients outside the Open Cowork / Claude ecosystem is not documented
  • No automated tests for this skill are evidenced in the repo; the guidance's correctness relies on user verification

How do you install this skill?

The skill ships with the Open Cowork desktop app in the repo's .claude/skills/skill-creator/ directory. To use it in another Agent Skills-compatible environment, copy the whole skill-creator folder into that environment's skills directory (e.g. .claude/skills/ for Claude Code). Desktop app installation: on macOS run brew tap OpenCoworkAI/tap && brew install --cask --no-quarantine open-cowork; on Windows download the .exe from GitHub Releases; the app supports adding and deleting custom skills. Standalone installation steps outside these environments are not documented in the source.

How do you use this skill?

  1. In a skill-triggering environment (Open Cowork or Claude Code), describe what you want, e.g. "Create a skill that turns monthly reports into PPT decks"; 2. Answer the clarification questions the skill asks; 3. Let Claude run the init, edit, and package steps in order; 4. The output is a <skill-name>.skill file you can share; 5. After real-world use, return to this skill to iterate on SKILL.md or bundled resources.

How does this skill compare with similar options?

The source positions it as the only meta-task skill among Open Cowork's five bundled skills (pptx, docx, pdf, xlsx, skill-creator): the other four perform concrete document processing, while this one teaches Claude to create new skills. Relative to Claude Cowork, Open Cowork is the open-source implementation, and skill-creator is its custom-skill development toolkit.

FAQ

Does creating a skill with this tool cost anything extra?
The skill is free under the MIT-licensed open-source project; you only pay for the AI model API usage you configure (Anthropic, OpenRouter, GLM, etc.).
Are the skills I create limited to my own machine?
The packaged output is a .skill file (zip format) that can be shared. However, the source does not document compatibility of the .skill format with other clients, so verify before cross-ecosystem distribution.
Can I use this without coding knowledge?
Open Cowork targets no-code usage and the creation flow is conversational; but if your skill needs executable scripts, packaging involves running Python scripts.
What if packaging fails?
package_skill.py reports validation errors and exits without creating a package; fix the reported frontmatter or structure issues and rerun. The skill also includes an explicit iteration step (Step 6) for ongoing improvements.

More skills from this repository

All from OpenCoworkAI/open-cowork

Related skills