Aug 25, 2026
awesome-claude-skills: A 10k-Star Cross-Repository Skill Index
A breadth-first curated list that organises a wide breadth of Claude Skills into 13 topical sections and links each entry back to its source repo, with a Collections section that surfaces multi-skill bundles like PM Skills and Agent Almanac.
There is no single place to discover Claude Skills. They live across individual repos, personal blogs, marketplace plugins, and aggregator sites. BehiSecc/awesome-claude-skills solves the discovery layer by curating the breadth of the ecosystem into one navigable index, with each entry linking back to its upstream repo.
Why This Skill Matters
This is not a skill you install. It is a cross-repository index. The README structures the catalogue into 13 topical sections (Document Skills, Development & Code Tools, Data & Analysis, Scientific & Research Tools, Writing & Research, Learning & Knowledge, Media & Content, Health & Life Sciences, Collaboration & Project Management, Security & Web Testing, Utility & Automation, Articles & Blog Posts, Collections), each grouping dozens of entries by functional domain. The Collections section then aggregates multi-skill bundles so you can adopt a curated pack instead of piecing skills together one at a time — PM Skills, Agent Almanac, agent-starter, AlterLab-Academic-Skills, goose-skills, Marketing Skills, and roughly two dozen more are listed by name in the README.
A tip callout in the README tells web-app builders "to do yourself a favor and use VibeSec-Skill to avoid getting hacked." Atlas Cloud is listed as the page sponsor.
If you need breadth rather than depth — when you are not yet sure which category of skill you want — this list is the fastest way to scan the space before committing to a stack.
Installation
This is a curated list, not an installable plugin. Clone or browse it, then follow each entry's link to its upstream repo where the actual install command lives.
git clone https://github.com/BehiSecc/awesome-claude-skills
cd awesome-claude-skills
There is no contribution command beyond the standard fork-and-PR flow described in the repo's Contribution section. Most entries reference install steps from their respective upstream projects rather than a unified command.
Real Workflow: Pick a Bundle for a Specific Job
Suppose you are joining a small team that needs Claude to help with three distinct jobs: pull quotes and stats from PDF reports, draft launch copy for product pages, and triage inbound support tickets. You could install skills one by one, or you could start from the Collections section and pick three that already group compatible skills.
Step 1. Open the README and jump to the Collections section. Scan for any bundle whose description mentions your three jobs. Concrete matches visible in the README: Marketing Skills (for product launch copy), career-ops and linkedin-skills (for outbound copy), and agentskill.sh (a discovery hub if the catalogue you want is not on this page).
Step 2. Click into Marketing Skills (or any candidate). The README does not install it for you — it links to the upstream repo. Open that repo and follow its install instructions.
Step 3. Verify the bundle is available in your agent. For Claude Code, a typical install follows the marketplace pattern:
/plugin marketplace add <upstream-org>/<bundle-repo>
/plugin install <bundle-name>@<upstream-org>
Step 4. Trigger one skill end-to-end and confirm it picks up the rest of the bundle's references. If the first invocation routes you to a specific reference, the bundle's routing is working.
The value of starting from this list is the breadth pass: you see 30+ bundles side by side and can pick the one whose description most closely matches your team's jobs.
Real Workflow: Vet a Skill Before Installing It
Suppose an entry in this list points to a skill you have never heard of, and you do not want to install it blind. The list itself does not run a vetting pass, but it gives you the inputs you need for one.
Step 1. From the list, click through to the upstream repo. Check three signals: star count (rough popularity proxy), last commit date (still maintained?), and the LICENSE file (compatible with your use?).
Step 2. Open two or three files referenced by the skill's SKILL.md — typically the ## When to Use and ## When Not to Use sections. If those sections are thin or missing, treat the skill as experimental.
Step 3. Read the entry's one-line description in this list again. If the description overpromises ("the only skill you need for X"), prefer another entry. The list's strength is curation, not endorsement — you still verify on the upstream side.
Step 4. Only after step 1–3 pass, copy the install command from the upstream repo and run it. The list never saves you from a bad install; it saves you from picking a bad skill in the first place.
- Start with the Collections section if you want a curated pack, or with a topical section (Development & Code Tools, Document Skills, etc.) if you know the job.
- When the README's one-line description is vague, click through — the upstream repo is the source of truth, not the list.
- For web apps, read the VibeSec-Skill tip callout in the README before installing skills that handle user input.
- Treat the Collections section as a hint, not an endorsement — verify each bundle's repo the same way you would any other skill.
- Open Issues (3) and Open PRs (90+) on the list itself indicate active curation; if a category feels stale, search the upstream repos directly.
When Not to Use This
If you already know the specific skill you want and have its repo URL, skip this list and install directly. It is a breadth-first discovery layer; it adds no value when the question is already answered. Also skip it for any skill that needs to run in a constrained environment where the upstream repo is blocked, since each install requires GitHub access anyway.
See the leaderboard for more curated Claude skill indexes.