hf-cli
by huggingfaceThe hf-cli skill helps you use the Hugging Face Hub CLI (`hf`) for authentication, downloads, uploads, repo and bucket management, dataset and model inspection, and other Hub workflows. It is useful for Backend Development teams that want repeatable, scriptable hf-cli usage and a practical hf-cli guide.
This skill scores 78/100, which means it is a solid directory candidate: users can likely trigger it reliably and get real Hugging Face CLI workflows without much guesswork. For directory users, it is worth installing if they work with Hugging Face Hub, auth, repos, jobs, datasets, Spaces, or endpoints, though they should expect a broad command surface rather than a narrow task-specific helper.
- Very strong triggerability: the description explicitly maps the skill to "hf", "huggingface", "huggingface-cli", and Hugging Face ecosystem tasks.
- Substantial operational coverage: it names downloading, uploading, auth, cache, repos, jobs, datasets, Spaces, webhooks, collections, and inference endpoints.
- No placeholder/demo signals: frontmatter is valid, body is substantial, and the repo shows concrete command-oriented content with repo/file references.
- No install command is embedded in SKILL.md, so users may need to rely on the command reference rather than a fully guided install flow.
- The skill is broad and CLI-heavy; users seeking a narrow workflow may need to read further to find the exact command path.
Overview of hf-cli skill
The hf-cli skill helps you use the Hugging Face Hub CLI, hf, to authenticate, download and upload files, manage repos and buckets, inspect models and datasets, and work with Hugging Face services from the terminal. It is a strong fit for Backend Development workflows that need repeatable, scriptable Hub access instead of one-off web clicks.
What hf-cli is for
Use the hf-cli skill when the task is operational: login state, cache handling, repo sync, dataset queries, endpoint setup, webhooks, jobs, or moving artifacts between local systems and the Hub. It is especially useful when a user already knows the Hugging Face ecosystem but needs the exact command flow, flags, and installation path.
When this skill is the right fit
Choose hf-cli if the goal is to automate Hub actions, integrate them into CI/CD, or standardize team workflows around the CLI. It is a better fit than a generic prompt when the user needs dependable command syntax, current auth behavior, or guidance on which hf subcommand matches the job.
What makes it different
The main value is practical command selection, not conceptual explanation. This hf-cli guide centers the modern hf command, notes that it replaces deprecated huggingface-cli, and helps users avoid guesswork around auth, cache, and Hub resource management.
How to Use hf-cli skill
Install and confirm the CLI
Install the skill with npx skills add huggingface/skills --skill hf-cli. Then confirm the CLI is available and current by checking hf --help and hf auth whoami. If you are migrating from older docs, treat huggingface-cli as legacy and prefer hf in new commands.
Turn your goal into a usable prompt
The best hf-cli usage starts with a concrete target, not a vague “help me with Hugging Face.” Include what you want to move or manage, where it lives, and any constraints. For example: “Upload a fine-tuned model folder to org/model-name, keep only config.json and model.safetensors, and authenticate with a token from CI.” That gives the skill enough context to choose the right subcommand and flags.
Read these files first
Start with SKILL.md, then inspect README.md, AGENTS.md, metadata.json, and any rules/, resources/, references/, or scripts/ folders if present. For this repo, SKILL.md is the primary source, so the main work is extracting the command model, supported tasks, and migration notes rather than chasing a large file tree.
Use the skill with real workflow constraints
Give the skill the same details you would give a teammate: repo ID, file paths, revision or branch, cache location, whether the command runs locally or in CI, and whether you need a dry run or minimal output. These inputs materially improve hf-cli install and hf-cli usage guidance because they narrow the command to the right resource and reduce accidental uploads, downloads, or auth mistakes.
hf-cli skill FAQ
Is hf-cli only for model downloads?
No. The hf-cli skill covers more than download and upload flows: authentication, cache management, repos, datasets, spaces, buckets, jobs, papers, and related Hub operations. If your work touches the Hugging Face ecosystem, hf-cli is often the correct starting point.
Do I need this if I already know shell commands?
Yes, if you want fewer command errors and faster setup. A normal prompt can explain the idea, but hf-cli is better when you need current CLI syntax, the right hf subcommand, or a migration path from deprecated huggingface-cli.
Is hf-cli good for beginners?
Yes, as long as the request is specific. Beginners usually get the best result by naming the task and the target repo, for example: “I need to log in and download one dataset snapshot for local testing.” That is easier to convert into a working command than a broad request like “show me Hugging Face CLI.”
When should I not use hf-cli?
Skip it if the task is purely conceptual, unrelated to the Hub, or better solved in the web UI with no automation need. It is also not the best fit if you only want general AI/ML advice without a CLI action.
How to Improve hf-cli skill
Give the hardest constraint first
The strongest hf-cli inputs include what must not happen: no full cache download, no overwrite, no public exposure, no interactive login, or no extra files. Constraints like these change the command choice and are often more important than the high-level goal.
Include the exact Hub object
Name the repo type and identifier: model, dataset, space, bucket, endpoint, or job. hf-cli output improves when the skill knows whether you are targeting org/repo, a specific revision, or a local directory that needs to be synced to the Hub.
Ask for the workflow, not just the command
If you want a usable hf-cli guide, ask for the command plus the shortest safe workflow: install, auth, verify, execute, and validate. That helps surface practical steps like hf auth whoami, cache checks, or revision selection that reduce trial and error.
Iterate with real output and errors
If the first command fails, paste the exact error, the command you ran, and the resource you targeted. That is the fastest way to improve hf-cli for Backend Development tasks because the next answer can correct flags, auth state, path assumptions, or Hub permissions instead of guessing.