skill

skill-creator

The meta-skill: scaffolds, validates, and teaches the design rules for building agent skills (Apache-2.0 mirror)

Dimerin1/skill-creatorcommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/Dimerin1/skill-creator itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 23, 2026

skill-creator

The meta-skill: a skill that builds other skills.

Agent skills are just a folder with a SKILL.md in it — which makes them easy to start and easy to get subtly wrong. The description doesn't trigger, the body dumps 2,000 words into every context, the scripts get rewritten from scratch on each run. This skill encodes the design rules and ships the scaffolding tooling.

What it does

Walks a six-step process — understand with concrete examples → plan reusable resources → scaffold → write → validate → iterate — and enforces the design principles that make skills actually work:

Progressive disclosure. Metadata is always in context, the body loads on trigger, bundled resources load only when needed. Keep SKILL.md under 500 lines and push detail into references/.
Degrees of freedom. Match specificity to fragility — prose for open-ended judgment, parameterized scripts for preferred patterns, rigid scripts for error-prone sequences.
Concision. The context window is a public good. Every paragraph has to justify its token cost.
The right bucket. scripts/ for deterministic code, references/ for docs read into context, assets/ for files used in the output. No READMEs, no changelogs, no setup guides inside a skill.

Tooling

Script	Purpose
`scripts/init_skill.py`	Scaffold a new skill folder — `SKILL.md` template, `agents/openai.yaml`, optional `scripts/`, `references/`, `assets/` dirs and example files
`scripts/generate_openai_yaml.py`	Generate or regenerate the UI metadata (`display_name`, `short_description`, `default_prompt`)
`scripts/quick_validate.py`	Check frontmatter, required fields, and naming rules

scripts/init_skill.py my-skill --path ~/.claude/skills \
  --resources scripts,references \
  --interface display_name="My Skill" \
  --interface short_description="Does the thing I keep redoing by hand" \
  --interface default_prompt="Do the thing"

scripts/quick_validate.py ~/.claude/skills/my-skill

Note that --interface is repeated per key — passing several key=value pairs after one flag is an argparse error. short_description must be 25–64 characters.

Install

PowerShell (Windows):

irm https://raw.githubusercontent.com/Dimerin1/skill-creator/main/install.ps1 | iex

bash (macOS/Linux):

curl -fsSL https://raw.githubusercontent.com/Dimerin1/skill-creator/main/install.sh | bash

Manual:

git clone https://github.com/Dimerin1/skill-creator.git
mkdir -p ~/.claude/skills/skill-creator
cp -r skill-creator/{SKILL.md,agents,references,scripts,assets,license.txt} ~/.claude/skills/skill-creator/
cp skill-creator/commands/skill-creator.md ~/.claude/commands/

Reload your Claude Code window afterwards, then type /skill-creator.

Known quirks

Carried over from upstream, documented rather than patched so this stays a faithful mirror:

The scaffolded SKILL.md has an unquoted description: [TODO: ...], which YAML parses as a list. Running quick_validate.py before you replace the TODO fails with Description must be a string, got list. Fill in the description first, or quote it.
SKILL.md points at references/workflows.md and references/output-patterns.md for workflow and output-format patterns. Neither ships in this bundle; only references/openai_yaml.md is present.
The text says "Codex" throughout — it applies unchanged to Claude Code.

License

Apache-2.0. license.txt is preserved from upstream; this repo is a mirror packaged for one-line install.