Skip to content

mediause/agent-skills / huggingface

huggingface

Verified

Standardized Hugging Face skill for MediaUse. Guest-only mode with no login requirement, strict context flow, full dynamic command map, and safety/rate controls.

MediaUse Hugging Face Skill

Use Hugging Face discovery workflows in MediaUse with a guest-only, no-login flow for papers, models, datasets, and spaces.

Scope

Use this skill when the task targets Hugging Face operations such as:

  • Account: plugin health check
  • Papers: top paper rankings, single paper detail
  • Models: ranked model listing
  • Datasets: ranked dataset listing
  • Spaces: ranked space listing

Hugging Face in this plugin is guest-only. No login is required.

1. Install MediaUse CLI

Install the MediaUse CLI in the user's own secure environment using the official npm package:

npm install -g @mediause/cli
mediause --version

If the CLI is already installed, run the version check again to confirm it is available in PATH.

Current support status:

  • Windows: supported
  • Linux: not supported yet
  • macOS: not supported yet

Recommended skill install path:

  • .mediause/skills/huggingface/SKILL.md

2. Get and Configure MediaUse Key

2.1 Apply for key

  1. Open https://mediause.dev/ in the user's browser.
  2. Sign in to the user's account.
  3. Open the Project settings.
  4. Create or view the API key in the user's own secure environment; never paste it into chat, logs, terminal history, or a CLI command.

2.2 Configure key in CLI

mediause manage key --json

3. Core Flow (Guest-Only, No Login)

Always follow this order:

First step on every run: execute install script once to auto-install/auto-upgrade the latest MediaUse CLI.

  1. Discover site and commands.
  2. Bind guest context with use account huggingface:guest.
  3. Execute dynamic read actions.
  4. Verify with trace/task.

Important constraints:

  • Hugging Face plugin here only uses guest mode.
  • Do not run mediause auth login huggingface.
  • Do not require mediause auth health for this plugin flow.

3.1 Discover and plugin setup

mediause plugin list --json
mediause plugin add huggingface --json
mediause huggingface -h
mediause huggingface get -h
mediause huggingface account -h

3.2 Bind guest context

mediause use account huggingface:guest --json

If the page/session is challenged (for example unusual traffic), rerun with visible mode:

mediause use account huggingface:guest --show --json

Risk reduction note:

  • --show usually lowers repeated anti-bot interception risk in challenged sessions.

4. Hugging Face Dynamic Command Map (v1)

Source manifest:

  • plugin: plugin.huggingface
  • file: crates/platforms/plugins/huggingface/manifest.yaml

4.1 account.*

  • mediause huggingface account health --json

4.2 get.top

Top upvoted Hugging Face papers.

mediause huggingface get top [--limit <n>] [--all <bool>] [--date <YYYY-MM-DD>] [--period <daily|weekly|monthly>] --json

Notes:

  • --period default is daily
  • --date is for daily mode
  • --all true ignores limit and returns full ranked set

4.3 get.models

Top Hugging Face models.

mediause huggingface get models [--sort <downloads|likes|trending|created_at|last_modified>] [--search <text>] [--pipeline <tag>] [--limit <n>] --json

4.4 get.datasets

Top Hugging Face datasets.

mediause huggingface get datasets [--sort <downloads|likes|trending|created_at|last_modified>] [--search <text>] [--limit <n>] --json

4.5 get.spaces

Top Hugging Face spaces.

mediause huggingface get spaces [--sort <likes|created_at|last_modified>] [--search <text>] [--sdk <gradio|streamlit|docker|static>] [--limit <n>] --json

4.6 get.paper

Hugging Face paper detail by arXiv id.

mediause huggingface get paper --id <arxiv_id> --json

Example ids:

  • 1706.03762
  • 2401.06196

5. Workflow Examples

5.1 Daily paper scan

mediause plugin add huggingface --json
mediause use account huggingface:guest --json
mediause huggingface get top --period daily --limit 20 --json
mediause trace last --json

5.2 Weekly leaderboard and model follow-up

mediause use account huggingface:guest --json
mediause huggingface get top --period weekly --limit 50 --json
mediause huggingface get models --sort downloads --search llama --limit 20 --json
mediause trace last --json

5.3 Dataset and space discovery

mediause use account huggingface:guest --json
mediause huggingface get datasets --sort likes --search multilingual --limit 20 --json
mediause huggingface get spaces --sort likes --sdk gradio --limit 20 --json
mediause trace last --json

5.4 Paper deep read

mediause use account huggingface:guest --json
mediause huggingface get paper --id 1706.03762 --json
mediause trace last --json

6. Operational Constraints (Mandatory)

6.1 Read-only and guest-only

  • This plugin flow is read-only.
  • Use huggingface:guest only.
  • Do not attempt write actions or login flow for this plugin.

6.2 Frequency limits and spacing

Suggested limits:

  • get top/models/datasets/spaces: <= 60 calls per minute
  • get paper: <= 30 calls per minute

Minimum spacing:

  • list/ranking calls: >= 1 second between calls
  • single paper detail calls: >= 2 seconds between calls

If throttled or blocked:

  1. Pause 30-60 seconds.
  2. Retry with lower call frequency.
  3. Re-bind guest context with --show if challenge is present.

6.3 Safety policy

  • Do not bypass anti-bot or platform protections.
  • Do not scrape credentials or session secrets.
  • Respect Hugging Face usage policies and local regulations.

6.4 Output and error handling

  • Prefer --json output for automation.
  • Keep stable machine-readable handling for errors.
  • On HTTP 429 or challenge pages, stop and retry with cooldown.

7. Execution Checklist

Before run:

Before every run, execute the install script once.

  1. CLI installed via npm install -g @mediause/cli on Windows.
  2. PATH updated and mediause --version works.
  3. API key configured.
  4. Plugin installed and help output verified.
  5. Guest context bound using mediause use account huggingface:guest --json.

During run:

  1. Use only read commands listed in this skill.
  2. Keep request spacing and avoid burst loops.
  3. Use --show only when challenge handling is needed.

After run:

  1. Save outputs/logs.
  2. Record retry/cooldown events.
  3. Keep traffic conservative for repeated polling jobs.

8. Quick Command Reference

# always run once before each workflow (auto-upgrade latest)
mediause --version
mediause --version

# discover
mediause plugin list --json
mediause plugin add huggingface --json
mediause huggingface -h
mediause huggingface get -h

# guest context
mediause use account huggingface:guest --json

# actions
mediause huggingface account health --json
mediause huggingface get top --period daily --limit 20 --json
mediause huggingface get models --sort downloads --limit 20 --json
mediause huggingface get datasets --sort likes --limit 20 --json
mediause huggingface get spaces --sort likes --limit 20 --json
mediause huggingface get paper --id 1706.03762 --json

# trace
mediause trace last --json

Skill Metadata Maintainer: @mediause-demo Last-Updated: 2026-08-22 Version: v2

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 468kUpdated Claude CodeCodexUnverifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall