Build reusable CLI scripts for Hugging Face API workflows

This skill builds reusable command-line scripts and utilities for the Hugging Face API, designed for chaining, piping, and intermediate processing - especially useful when combining multiple API calls or when a task will be repeated or automated. It works against both the raw HTTP API and the hf CLI tool, and can read model/dataset cards directly from repositories. It enforces a set of script rules: every script takes --help describing its inputs and outputs, non-destructive scripts are tested before handoff, shell scripts are preferred unless complexity calls for Python or TSX, the HF_TOKEN environment variable is always passed as an Authorization: Bearer header for higher rate limits and proper access, the shape of API results is investigated before committing to a design, and usage examples are shared once a script is complete. It ships reference example scripts covering common patterns: hf_model_papers_auth.sh chains trending models to model metadata to model-card parsing with fallbacks while handling auth automatically; find_models_by_paper.sh does authenticated search with a retry path for overly narrow arXiv-prefixed queries; hf_model_card_frontmatter.sh uses the hf CLI to download model cards and emit NDJSON summaries (license, pipeline tag, tags, gated-prompt flag); baseline ultra-simple examples exist in bash, Python, and TypeScript for raw JSON output; and hf_enrich_models.sh reads model IDs from stdin and streams enriched NDJSON, composable with other scripts via piping (e.g. sorting the top-10 most-downloaded models from a baseline query). It documents the high-level API endpoints available at huggingface.co (/api/datasets, /api/models, /api/spaces, /api/collections, /api/daily_papers, /api/notifications, /api/settings, /api/whoami-v2, /api/trending, /oauth/userinfo), and the OpenAPI spec at /.well-known/openapi.json - explicitly warning not to read that file directly since it is too large, and to query it through jq instead. It also documents the hf CLI's full command surface (auth, buckets, cache, collections, datasets, discussions, download, endpoints, env, extensions, jobs, models, papers, repos, skills, spaces, sync, upload, upload-large-folder, version, webhooks), noting it has replaced the deprecated huggingface-cli.

Use it when building a tool or script to accomplish a task using Hugging Face API data - especially when chaining or combining multiple API calls, or when the task will be repeated or automated. Confirm user preferences where there are open questions before finalizing a script's design. Do NOT attempt to read the full openapi.json spec directly - it's documented as too large to process - always filter it through jq first, and constrain any exploratory API queries to small result counts to keep them easy to process while staying representative.

Inputs: the task or automation goal, and (for authenticated calls) an HF_TOKEN. Outputs: a reusable, --help-documented script (shell, Python, or TSX) that fetches, enriches, or chains Hugging Face API/CLI data, tested for non-destructive behavior and shared with usage examples once complete.

curl -s "https://huggingface.co/.well-known/openapi.json" | jq '.paths | keys | sort'

Built on the Hugging Face HTTP API (documented via OpenAPI at /.well-known/openapi.json) and the hf CLI tool (successor to the deprecated huggingface-cli), composed with standard Unix tooling like jq and curl for piping and chaining.

Developers who need a reusable, scriptable way to fetch, enrich, or chain Hugging Face Hub data - models, datasets, papers, spaces - for a repeated or automated workflow rather than a one-off manual query.