메뉴
File and git-like operations on HuggingFace repositories (model / dataset / space). Covers list / read / upload / delete of files, branches, tags, pull requests, and repo management (create / update). Use to inspect repos, ship artifacts, and propose changes via PRs.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Alternative paper search backends — OpenAlex (~250M works, no auth, strong PDF extraction) and CORE (~49M full-text, BODY search). Use when hf-papers can't find a paper, when Semantic Scholar is rate-limited, when you need PDFs of older/long-tail papers, or when you need to search the actual body text of papers (not just titles/abstracts).
Discover GitHub repositories, find example scripts within them, and read source files with line-range support and automatic Jupyter notebook conversion. Use BEFORE writing any ML training, fine-tuning, or inference code — your knowledge of library APIs is outdated and example scripts show the current patterns.
Inspect a HuggingFace dataset in one call — status, configs/splits, schema, sample rows, parquet file listing, and detailed messages-column analysis (roles, keys, tool-calls). REQUIRED before submitting any training job to verify column format matches the training method.
Explore and fetch HuggingFace library documentation (transformers, trl, datasets, peft, accelerate, trackio, vllm, optimum, courses, etc.). Use when you need current trainer arguments, config parameters, or API reference that your internal knowledge may have wrong.
Submit and manage HuggingFace compute jobs — run training / fine-tuning / inference / batch work on HF cloud infrastructure in Python or Docker mode. Supports one-shot and recurring (cron-style) jobs.
Discover ML research papers, read their contents by section, trace citations, find linked datasets/models — combining HuggingFace Papers, arXiv, and Semantic Scholar. Use whenever you need to research a method, find training recipes with published results, crawl a citation graph, or locate the datasets a paper used. This is your starting point for any literature-first workflow.
| name | hf-repos |
| description | File and git-like operations on HuggingFace repositories (model / dataset / space). Covers list / read / upload / delete of files, branches, tags, pull requests, and repo management (create / update). Use to inspect repos, ship artifacts, and propose changes via PRs. |
Two scripts:
scripts/hf_repo_files.py — file operations: list, read, upload, delete (with wildcard patterns).scripts/hf_repo_git.py — git-like operations: branches, tags, PRs, repo create/update.config.json, README.md, tokenizer configs) on a model/dataset repo.hf-jobs for the training run itself; this is for the push).hf_repo_files.py (file ops)# Peek at a model repo
python scripts/hf_repo_files.py list --repo-id meta-llama/Llama-2-7b
# Read a config
python scripts/hf_repo_files.py read --repo-id gpt2 --path config.json
# Upload a README
python scripts/hf_repo_files.py upload \
--repo-id me/my-model --path README.md \
--content-file ./README.md \
--commit-message "Initial README"
# Upload as a PR (propose, don't commit)
python scripts/hf_repo_files.py upload \
--repo-id org/model --path fix.py \
--content-file ./fix.py --create-pr
# Delete with wildcards
python scripts/hf_repo_files.py delete \
--repo-id me/my-model --pattern '*.tmp' --pattern 'logs/'
Flags:
--repo-type {model|dataset|space} (default model).--revision (branch / tag / commit, default main).--content-file (file path), --content-stdin (pipe), or --content (inline string).hf_repo_git.py (git-like ops)Branch / tag flow:
python scripts/hf_repo_git.py list-refs --repo-id me/my-model
python scripts/hf_repo_git.py create-branch --repo-id me/my-model --branch experiment-v2
python scripts/hf_repo_git.py create-tag --repo-id me/my-model --tag v1.0 --revision main
PR flow (HF uses "discussions" as PRs):
# 1. Create draft PR (empty by default)
python scripts/hf_repo_git.py create-pr --repo-id org/model --title "Fix tokenizer config"
# 2. Add commits via file upload, targeting the PR ref
python scripts/hf_repo_files.py upload \
--repo-id org/model --path tokenizer_config.json \
--content-file ./fix.json \
--revision refs/pr/1
# 3. Publish the draft (draft → open)
python scripts/hf_repo_git.py change-pr-status --repo-id org/model --pr-num 1 --new-status open
# 4. Merge (for your own repo)
python scripts/hf_repo_git.py merge-pr --repo-id me/my-model --pr-num 1
# Comment or close
python scripts/hf_repo_git.py comment-pr --repo-id me/my-model --pr-num 1 --comment "LGTM"
python scripts/hf_repo_git.py close-pr --repo-id me/my-model --pr-num 1
Repo management:
python scripts/hf_repo_git.py create-repo --repo-id me/my-new-model --private
python scripts/hf_repo_git.py create-repo --repo-id me/my-space --repo-type space --space-sdk gradio
python scripts/hf_repo_git.py update-repo --repo-id me/my-model --gated auto
python scripts/hf_repo_git.py update-repo --repo-id me/my-model --private false
These mutate published state and should be confirmed with the user:
delete-branch, delete-tag, merge-pr, create-repo, update-repo, hf_repo_files.py delete, hf_repo_files.py upload (can overwrite files).The scripts don't gate them — your harness should. When in auto mode, think twice.
HF_TOKEN — required.hf-jobs — the other half of the deploy loop: train on Jobs, push artifacts with this.hf-datasets — for deep dataset inspection, prefer hf-datasets over hf_repo_files.py read on raw parquet.