huzey/claude-skills

---
dataset_info:
features:
- name: name
dtype: large_string
- name: description
dtype: large_string
- name: full_content
dtype: large_string
- name: repo
dtype: large_string
- name: split
dtype: large_string
- name: qwen3emb_description
list: float64
- name: gpt_domain
dtype: large_string
- name: gpt_capability
dtype: large_string
- name: gpt_use_case
dtype: large_string
- name: gpt_triggers_constraints
dtype: large_string
- name: domain_category
dtype: large_string
- name: domain_subcategory
dtype: large_string
- name: qwen3emb_full_content
list: float32
- name: qwen3emb_gpt_domain
list: float32
- name: qwen3emb_gpt_capability
list: float32
- name: qwen3emb_gpt_use_case
list: float32
- name: qwen3emb_gpt_triggers_constraints
list: float32
- name: ref_files
dtype: large_string
- name: skills_sh_id
dtype: string
- name: skills_sh_total_installs
dtype: int64
- name: skills_sh_weekly_installs
dtype: int64
- name: github_stars
dtype: int64
splits:
- name: train
num_examples: 22862
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---

# Claude Skills Dataset

This dataset contains curated SKILL.md files plus generated structured summaries and embeddings.

## Columns

- `name`: skill name (from SKILL.md frontmatter)
- `description`: short description (from SKILL.md frontmatter)
- `full_content`: full SKILL.md content (includes frontmatter metadata)
- `repo`: source repository
- `split`: dataset split label
- `qwen3emb_description`: embedding vector for `description` (float list)
- `gpt_domain`: concise domain label extracted from the skill content
- `gpt_capability`: one-sentence capability summary
- `gpt_use_case`: one-sentence use-case summary
- `gpt_triggers_constraints`: triggers/constraints if present, else empty string
- `domain_category`: broad rule-based category from `gpt_domain` (tokenized, rule-based)
- `domain_subcategory`: coarse subcategory within `domain_category` (rule-based; `Other` fallback)
- `qwen3emb_full_content`: embedding vector for `full_content` after stripping YAML frontmatter (`--- ... ---`)
- `qwen3emb_gpt_domain`: embedding vector for `gpt_domain`
- `qwen3emb_gpt_capability`: embedding vector for `gpt_capability`
- `qwen3emb_gpt_use_case`: embedding vector for `gpt_use_case`
- `qwen3emb_gpt_triggers_constraints`: embedding vector for `gpt_triggers_constraints`
- `ref_files`: inlined reference docs from the skill folder (path + `---` + content blocks), may be empty
- `skills_sh_id`: canonical `skills.sh` id guess (`{repo}/{slugified(name)}`)
- `skills_sh_total_installs`: installs count from skills.sh search index (leaderboard "Installs")
- `skills_sh_weekly_installs`: weekly installs from the skills.sh skill page ("Weekly Installs")
- `github_stars`: GitHub stars shown on the skills.sh skill page (repo-level; repeated across skills in same repo)

## Source

- Crawled from skills.sh, 2026/04/02. Enriched with skills.sh installs + GitHub stars, 2026/04/06.