# Manage Hugging Face Hub Resources with hf

Hugging Face Hub operations span many commands, resources, and safety considerations. This skill selects suitable `hf` workflows and explains their effects.

## Install

```bash
npx skillstore add sickn33/hugging-face-cli
```

## Metadata

- Status: approved
- Slug: sickn33-hugging-face-cli
- Skillstore revision: r2
- Version status: missing
- Tree hash: 09d48f895f6cb5c4f7937337960f5c3e76c2ccf20f8f02e42e1ff8e285fd1bb9
- Author: sickn33
- GitHub username: sickn33
- License: Apache-2.0
- Repository: https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/hugging-face-cli
- Ref: 88a8e9a07f4c54ab105c1c41b6267c287146b07b
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 79
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/sickn33-hugging-face-cli
- Manifest: https://skillstore.pages.dev/api/skills/sickn33-hugging-face-cli/manifest

## Capabilities

- Selects \`hf\` commands for authentication, downloads, uploads, cache management, and environment checks.
- Guides management of models, datasets, repositories, collections, discussions, Spaces, buckets, and webhooks.
- Plans remote jobs, scheduled jobs, SSH access, logs, labels, hardware, and resource settings.
- Explains endpoint deployment, updates, scaling, pausing, resuming, and deletion.
- Supports machine-readable, quiet, and human-readable output choices for automation workflows.
- Documents CLI extensions, assistant skills, and \`hf-mount\` repository or bucket mounts.

## Use Cases

- Transfer model artifacts: Choose safe download or upload options, revisions, filters, destinations, and output formats for model repositories.
- Operate hosted workloads: Plan jobs, schedules, endpoints, Spaces, hardware settings, logs, and lifecycle actions with approval checkpoints.
- Explore Hub resources: Search models, datasets, papers, leaderboards, collections, and repository metadata using focused CLI queries.

## Prompt Templates

### Check account access

```
Confirm which Hugging Face account is active and explain the result. Do not print access tokens.
```

### Download a model revision

```
Prepare the safest `hf` command to download [REPO_ID] at [REVISION] into [LOCAL_DIR]. Explain authentication and disk requirements.
```

### Plan a bucket synchronization

```
Create a dry-run-first plan to synchronize [LOCAL_DIR] with bucket [BUCKET_ID]. Preserve remote files unless I explicitly approve deletion.
```

### Deploy and operate an endpoint

```
Design an `hf` CLI workflow to deploy [MODEL_REPO] as endpoint [NAME] in [REGION]. Include cost checks, secret handling, verification, and rollback.
```

## Limitations

- Requires an installed \`hf\` CLI, network access, and suitable Hugging Face credentials.
- The command catalog was generated with \`huggingface\_hub\` 1.21.0 and may differ from later releases.
- Does not determine current pricing, quotas, hardware availability, or organization policies.
- Destructive, costly, public, and credential-related operations still require explicit user review.

## Best Practices

- Use dry runs, saved plans, and explicit confirmations before deletion, synchronization, deployment, or public visibility changes.
- Keep tokens in approved credential storage or environment variables, and never print or place them in command history.
- Pin repository revisions and verify current CLI help, pricing, quotas, and hardware availability before execution.

## Anti Patterns

- Do not run delete, merge, sync-delete, or infrastructure commands without reviewing the exact target and impact.
- Do not pass access tokens directly in commands, logs, prompts, or shared files.
- Do not install extensions or remote scripts without reviewing their source, pinning versions, and verifying integrity.

## Security Audit

- Audited at: 2026-07-24T00:28:05.163\+00:00
- Summary: Review confirmed two medium network risks in the documented installers: each fetches a mutable remote script that is later executed after manual inspection. The remaining 187 findings are false positives caused by Markdown code spans, metadata links, controlled temporary-directory handling, and ordinary CLI documentation. No prompt injection, credential exfiltration, or malicious intent was found.

## Stats

- Views: 105
- Downloads: 9
- Favorites: 0
- Popularity score: 0