Skip to content
Projects

vantaflow/claude-skills-collection

View on GitHub

vantaflow/claude-skills-collection has 30 GitHub stars and 9 forks. GitHot measured 0 stars in 24 hours and 0 in 7 days. Last checked 2026-11-22 13:11 UTC. Last push 2026-10-18.

How these counts are measured

Measured star growth

in 24h
0
in 7 days
0
in 15 days
Collecting history
in 1 month
Collecting history

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

About

349 skills open-source pour Claude Code — 34 domaines : dev, agents IA, AI/ML, cloud, database, DevOps, testing, sécurité, networking, Linux, automation, marketing, IoT, communication, data, freelance, management, et vie quotidienne.

Productivity

README.en.md

Original

🧠 Claude Skills Collection

The largest open-source skill collection for Claude — 348 skills across 34 domains: development, security, AI agents, DevOps, data, business, and everyday life.

A skill turns Claude into a specialised assistant with a structured, step-by-step workflow.

SkillsCategories LicenseBundles

🇫🇷 Version française — the canonical, most detailed document.


⚠️ Read this first: the skills are written in French

The body of every skill is French. The trigger phrases are bilingual: each description ends with an Also triggers on "…" segment listing English phrasings, so an English prompt routes to the right skill. Claude then answers you in the language you wrote in — the French body is guidance for the model, not output shown to you.

If you only ever prompt in English, this collection still works. If you want the skills themselves translated, that's an open contribution — see Contributing.


📥 Installation

1. A single bundle — recommended

The collection ships as seven thematic plugins. Each one installs on its own and loads only its own skills.

/plugin marketplace add https://github.com/vantaflow/claude-skills-collection
/plugin install claude-skills-security     # 10 skills, ~1,800 tokens of context
Plugin Domain Skills Standing context
claude-skills-dev Development & testing 120 ~18,600 tok
claude-skills-agents AI agents, LLM & prompting 68 ~11,600 tok
claude-skills-cloud-ops Cloud, DevOps & networking 40 ~5,900 tok
claude-skills-data Data & databases 19 ~2,700 tok
claude-skills-security Security 10 ~1,800 tok
claude-skills-business Business, career & writing 39 ~6,200 tok
claude-skills-life Health, wellbeing & daily life 64 ~11,700 tok
claude-skills-collection Everything 348 ~56,400 tok

Why the context column matters. The name and description of every installed skill are injected into every session you start, whether the skill fires or not. Installing all 348 costs roughly 56,000 standing tokens. One bundle costs a tenth of that. Pick the domain you actually work in. The figures come from claude plugin details and are recomputed by npm run check at each build.

2. The whole collection

/plugin marketplace add https://github.com/vantaflow/claude-skills-collection
/plugin install claude-skills-collection

All 348 skills become slash commands (/dev-docker-composer, /agent-spawner, /cloud-aws-architect, /security-threat-modeling…).

3. À la carte — one skill, or one category

Lightest option: only the skills you pick are copied into ~/.claude/skills.

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/vantaflow/claude-skills-collection/main/install.sh | sh -s -- dev-code-reviewer --launch
curl -fsSL .../install.sh | sh -s -- --category security     # every security skill
curl -fsSL .../install.sh | sh -s -- --search redis          # search before installing
curl -fsSL .../install.sh | sh -s -- --list                  # categories and volumes

# Windows (PowerShell)
iex "& { $(iwr -useb https://raw.githubusercontent.com/vantaflow/claude-skills-collection/main/install.ps1) } dev-code-reviewer -Launch"

📖 Online catalogue & manuals: vantaflow.github.io/claude-skills-collection/manuals — one manual per skill, instant search, one-command install. 🤖 Machine-readable index: skills.json (name, category, summary, triggers, paths) for your own tooling.


🎯 How a skill fires

You don't have to type the slash command. Describe what you need and the matching skill activates:

"review this PR for security issues"      → /dev-code-reviewer
"my Postgres query takes 8 seconds"       → /database-query-optimizer
"design a multi-agent research system"    → /agent-multi-agent-architect
"write a threat model for this API"       → /security-threat-modeling

Not sure which one? /meta-workflows reads your request and routes you to the most relevant skills in the catalogue.


🏗️ Architecture & build

  • Single source of truth: the <category>-skills/, docs/ and meta-skills/ directories (short kebab-case names). Edit skills there.
  • Generated artefacts — never hand-edit: skills/ (plugin payload), manuals/ (site), docs/SKILL_CATALOG.md, skills.json, and .claude-plugin/marketplace.json. Every skill in the plugin is prefixed by its category (dev-skills/docker-composer → dev-docker-composer) so slash commands can't collide.
  • A domain-appropriate Communication Rules block is appended at build time (terse for technical domains; supportive with a disclaimer for human ones).
npm run check          # strict frontmatter, collisions, lengths, links, counters, context cost
npm run check:routing  # two skills competing for the same trigger words
npm test               # build-script tests
npm run build          # check + routing + skills/ + manuals/ + catalogue + marketplace

npm run check:routing fails when two skills share two or more quoted trigger phrases, or carry the same directory name in two categories — a prompt containing those words would be routed arbitrarily. Legitimate homonyms are declared, with a written reason, in scripts/routing-allowlist.json.

⚠️ .claude-plugin/plugin.json must not exist. The marketplace.json entries are strict: false and carry all metadata; the mere presence of a root plugin.json breaks loading for every plugin with "conflicting manifests". The build aborts if the file reappears.

CI (validate.yml) replays check + routing + tests + build and fails if a committed artefact is stale.


🤝 Contributing

  • Full guide: CONTRIBUTING (French).
  • A skill fires wrongly or says something false: open an issue. The exact prompt you typed is the single most useful piece of information.
  • Proposing a skill: check the catalogue first — with 348 skills, yours may be an improvement to an existing one.
  • Translating skills to English is welcome. Keep the bilingual trigger convention (Se déclenche avec "…" + Also triggers on "…").
  • Security: SECURITY.md. Report vulnerabilities as a private advisory, never as a public issue.

Two rules that fail CI and are easy to forget: editing a generated artefact instead of the source, and writing a description whose triggers overlap an existing skill.


📄 License

MIT — see LICENSE. Use, modify and redistribute freely, including commercially.


⚠️ Disclaimers

Skills in the health, wellbeing, legal, finance and parenting categories are support tools, never professional advice. They consistently point you back to a qualified professional. Security and pentest skills carry their own scope rules: written authorisation, defined perimeter, responsible disclosure. Do not use them outside that frame.

Explore related projects

Loading…