datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
skillcenter-ir
SkillCenter Intent IR retrieval corpus
This is the CID-keyed retrieval release at
Publicus/skillcenter-ir. It
converts the complete
Tommysha/skillcenter-bundles
SkillCenter corpus and its local retrieval artifacts into
thin-client-friendly, Zstandard-compressed Parquet. It is bound to upstream
revision f9dd4fec3c86d85ebf116c7408ac5ce602c418a1 and contains:
216,972 canonical skills keyed by entry_cid;
3,776,520 BM25 terms and
107,971,682 document-term postings;
434,135 graph… See the full description on the dataset page: https://huggingface.co/datasets/Publicus/skillcenter-ir.SkillMD-138K
SkillMD-138K
A large public collection of Agent Skill files (SKILL.md) for empirical research.
Overview
Metric
Value
Total skills
138,133
Distinct repositories
20,556
Deduplicated
Yes (SHA-256 content hash)
What are Agent Skills?
Agent Skills are modular instruction files (typically named SKILL.md) that extend LLM agent capabilities without fine-tuning. Each skill contains YAML frontmatter (routing metadata) and a Markdown body… See the full description on the dataset page: https://huggingface.co/datasets/FayeZC/SkillMD-138K.SkillsBench-1650
SkillBench-1650
A benchmark dataset for evaluating safety detection systems on AI agent skill packages. Contains 1,500 benign and 150 malicious Claude Code skill samples, designed for multi-dimensional security scoring research.
Dataset Summary
Split
Samples
Description
benign
1,500
Real-world skills sampled from open-source repositories
malicious
150
Adversarial payloads injected into real skill hosts
Total
1,650
Benign samples are sourced from… See the full description on the dataset page: https://huggingface.co/datasets/zenith6888/SkillsBench-1650.droid_merged_skills_moving
Merged DROID Skill Dataset: moving
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_moving
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/moving
Skill: moving
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_moving",
"jellyho/droid_move_banana"
]
source_dataset
episodes… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_moving.droid_merged_skills_wiping
Merged DROID Skill Dataset: wiping
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_wiping
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/wiping
Skill: wiping
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_wiping",
"jellyho/droid_wipe_table"
]
source_dataset
episodes… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_wiping.claude-skills
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… See the full description on the dataset page: https://huggingface.co/datasets/huzey/claude-skills.skill-diffs
skill-diffs
Commit-by-commit revision history of agent skills (SKILL.md files) scraped from public GitHub repos. Each record is a (before, after, intent) tuple capturing how a skill was iteratively refined through human feedback.
v0.5 covers 4 platforms — Anthropic Claude, OpenClaw, OpenCode, and Hermes Agent — with PR title/body metadata as richer intent labels, MinHash + semantic clustering for dedup, structural diff_summary for filtering by edit type, aggregate quality_score for… See the full description on the dataset page: https://huggingface.co/datasets/shl0ms/skill-diffs.droid_merged_skills_stacking
Merged DROID Skill Dataset: stacking
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_stacking
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/stacking
Skill: stacking
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_stacking",
"jellyho/droid_stack_bar"
]
source_dataset
episodes… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_stacking.droid_merged_skills_picking
Merged DROID Skill Dataset: picking
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_picking
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/picking
Skill: picking
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_picking",
"jellyho/droid_pick_carrot"
]
source_dataset
episodes… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_picking.agent-skills-security-grades
Agent Skills Security Grades
Security grades and quality scores for 130,173 open-source AI agent skills and
MCP servers collected from GitHub, from Agent Skills Hub.
Each row is one skill/server with a rule-based security grade
(SAFE / CAUTION / UNSAFE / REJECT / UNAUDITED), red-flag identifiers, and a
0–100 quality score.
Why this exists
AI coding agents install third-party skills that run with the agent's full
permissions and credentials, but marketplaces rank… See the full description on the dataset page: https://huggingface.co/datasets/jasonzhuyansen/agent-skills-security-grades.droid_merged_skills_scooping
Merged DROID Skill Dataset: scooping
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_scooping
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/scooping
Skill: scooping
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_scooping",
"jellyho/droid_scoop_candy"
]
source_dataset… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_scooping.droid_merged_skills_pressing
Merged DROID Skill Dataset: pressing
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_pressing
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/pressing
Skill: pressing
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_pressing",
"jellyho/droid_press_button"
]
source_dataset… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_pressing.eval_skill-set-r1_expt4_place-upperThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 10,
"total_frames": 8257,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/eval_skill-set-r1_expt4_place-upper.roles-based-on-skillssim_mug_skillgen_v2This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"observation.state": {
"dtype": "float32",
"shape": [
10
],
"names": [
"x",
"y",
"z",
"r6_0",
"r6_1",
"r6_2",
"r6_3",
"r6_4"… See the full description on the dataset page: https://huggingface.co/datasets/Sounderya/sim_mug_skillgen_v2.droid_merged_skills_placing
Merged DROID Skill Dataset: placing
This dataset merges a language-filtered DROID skill subset with newly collected
environment data for the same skill.
Repository
Hugging Face repo: jellyho/droid_merged_skills_placing
Local merged dataset path: /scratch/jellyho/droid_merged_skills_v21/placing
Skill: placing
LeRobot codebase version: v2.1
Sources
[
"jellyho/droid_subsets_placing",
"jellyho/droid_place_hammer"
]
source_dataset
episodes… See the full description on the dataset page: https://huggingface.co/datasets/jellyho/droid_merged_skills_placing.pii-skills-ablation-results
PII Skills Ablation — Scored Results
This repository contains model predictions and evaluation scores for the ablation study described in:
"Asymmetry, Not Capability: Evaluation Shapes Tool-Augmented PII Detection in Small Language Models"
Results are produced by running four open-weight instruction-tuned models (Gemma 2 9B, Llama 3.1 8B, Mistral 7B, Qwen 2.5 7B) under four primary conditions (zero-shot, +Docs, +Tool, +Skills), plus three baselines (standalone PII-Codex detector… See the full description on the dataset page: https://huggingface.co/datasets/EdyVision/pii-skills-ablation-results.skill-set-r1This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 240,
"total_frames": 209337,
"total_tasks": 5,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:240"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/skill-set-r1.rollout_longhorizon_chain_skill2_1786411332_20260810_182344This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6… See the full description on the dataset page: https://huggingface.co/datasets/srinivas2002/rollout_longhorizon_chain_skill2_1786411332_20260810_182344.rollout_longhorizon_chain_skill1_1786411332_20260810_182222This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6… See the full description on the dataset page: https://huggingface.co/datasets/srinivas2002/rollout_longhorizon_chain_skill1_1786411332_20260810_182222.skill-set-r1_valThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 48,
"total_frames": 41270,
"total_tasks": 2,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:48"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/skill-set-r1_val.skill_align
SkillsAlign
Anonymous submission to NeurIPS 2026 — Datasets & Benchmarks Track.
Camera-ready metadata will be added after acceptance.
A 245 K-skill in-the-wild dataset for cross-context misalignment in AI
agent skill packages — when a skill's metadata description does not match
its instruction body or the resource files it ships. The dataset supports
training and evaluation of detectors that flag misaligned packages before
an agent loads them.
Code, Docker pipelines, and the paper… See the full description on the dataset page: https://huggingface.co/datasets/anon-skillsalign-26/skill_align.skill-leaderboard-testclaude-skills-diff
claude-skills-diff
This dataset contains Git history snapshots for SKILL.md files from repos in huzey/claude-skills.
What is a row?
Each row corresponds to a (repo, skill_path) pair where the content changed a lot between its first and last commit for that path.
We export:
text_initial / initial_sha / initial_date
text_final / final_sha / final_date
optional text_middle / middle_sha / middle_date
diffs: diff_initial_final and if middle exists: diff_initial_middle… See the full description on the dataset page: https://huggingface.co/datasets/huzey/claude-skills-diff.eval_skill-set-r2_expt1_close-upperThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 10,
"total_frames": 8258,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/eval_skill-set-r2_expt1_close-upper.all-models-cv-skill-jd-req-matchskill-set-splitedThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 244,
"total_frames": 213471,
"total_tasks": 6,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:244"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/skill-set-splited.Skillex-research
Skillex Research
Skillex Research is a deterministic, training-ready transformation of
Anthropic/enabling-independent-research
for the Gram/Skillex backend. It converts the upstream privacy-preserving aggregate cluster tables into:
clusters: a normalized evidence and retrieval view with stable identifiers, explicit attribution,
confidence intervals, sparse cross-facet signals, and deterministic train/validation/test splits;
gram_skillex: paired supervised examples for… See the full description on the dataset page: https://huggingface.co/datasets/Cenedril/Skillex-research.eval_skill-set-r1_expt4_close-lowerThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 10,
"total_frames": 8257,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/eval_skill-set-r1_expt4_close-lower.eval_skill-set-r3_expt1_place-lowerThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 10,
"total_frames": 8258,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/KS325/eval_skill-set-r3_expt1_place-lower.