35,802 agent skills, subagents, slash commands, prompts and rules files curated from 186 GitHub repositories β browsable, greppable, installable.
See the code35,802 ready-to-use skills, subagents, slash commands, prompts and rules β harvested, deduplicated and organized from 186 repositories across the Claude Code / Cursor / Codex / Copilot skill ecosystem, so you never have to write one from scratch again.
Browse the index Β· Search the catalog Β· Quick start Β· Credits Β· Contribute
AI coding agents (Claude Code, Cursor, Codex, Copilotβ¦) got a superpower this year: Skills β small, self-contained instruction packs that teach an agent how to do one thing well (write a spec, review a PR, design an API, run a security audit, format a resume). The problem: they're scattered across hundreds of individual GitHub repos, half-documented, and impossible to search in one place.
This repo fixes that. Every skill-bearing repo in the ecosystem, pulled together, deduplicated, categorized, and made greppable in one file. Whether you're a student who just discovered agent skills exist, or a staff engineer wiring up a fleet of subagents in production β the same catalog works for you.
| Kind | Count | Where | What it is |
|---|---|---|---|
| π§© Agent Skills | 18,681 | library/ | Full SKILL.md bundles (+ scripts, templates, references) |
| π€ Subagents | 2,888 | agents/ | Specialized agent role definitions |
| β‘ Slash Commands | 2,773 | commands/ | /command shortcuts |
| π¬ Prompts | 10,074 | prompts/ | System prompts, prompt corpora, leaked prompts |
| π Rules | 1,386 | rules/ | CLAUDE.md / AGENTS.md / .cursorrules exemplars |
| π Reference | 543 | reference/ | Awesome-lists, cheatsheets, handbooks (verbatim) |
git clone https://github.com/thedixitjain/the-mega-skill-library.git
Pick the path that matches how much you already know:
Copy the router skill. It searches the other 35,801 assets for you β nothing else loads into context or slows your agent down.
cp -r the-mega-skill-library/master-skill ~/.claude/skills/master-skill-library
Now just talk to Claude Code normally. When a task matches something in the library, it finds the right skill and uses it automatically. That's it β you're done.
Claude Code reads every installed skill's frontmatter at startup, so only install what you'll actually use β dozens, not thousands.
cp -r the-mega-skill-library/library/engineering-core/tdd-guide ~/.claude/skills/
Project-only instead of global? Copy into <project>/.claude/skills/ instead
of ~/.claude/skills/.
grep -i "pdf" CATALOG.tsv
awk -F'\t' '$2=="security-and-compliance" && $1=="skill"' CATALOG.tsv
grep -i "code review" CATALOG.tsv | awk -F'\t' '$1=="agent"' | cut -f3,6
CATALOG.tsv is one row per asset: kind Β· category Β· name Β· path Β· source_repo Β· description.
Most files here are plain Markdown with YAML frontmatter β portable by design.
library/ bundles follow the Agent Skills spec (works with any harness that
supports it); rules/ files map directly to .cursorrules / AGENTS.md /
custom instructions; prompts/ and commands/ are copy-paste text you can
drop into any system-prompt or slash-command config.
22 categories, auto-sorted by asset count. Click through for the full browsable table of every skill/agent/command/prompt in that category.
| Category | Assets | Covers |
|---|---|---|
| π¦ General purpose | 5,575 | Uncategorized / cross-cutting |
| π€ AI agents & harness | 4,583 | Agent design, orchestration, subagents, context engineering |
| π¬ Prompt engineering | 2,868 | Prompt techniques, system prompts, optimizers, persona design |
| βοΈ Engineering core | 2,805 | Refactoring, debugging, review, architecture, git, TDD |
| π¬ Research & academic | 2,483 | Literature review, papers, journals, grants, regulatory science |
| βοΈ DevOps & infra | 1,605 | Docker, K8s, Terraform, CI/CD, cloud, observability, SRE |
| π Security & compliance | 1,604 | AppSec, pentest, OSINT, threat modeling, SOC2/ISO/GDPR |
| π£ Marketing & growth | 1,584 | SEO/AEO, ads, CRO, social, campaigns, growth loops |
| ποΈ Backend & data | 1,583 | APIs, databases, schemas, queues, caching, pipelines |
| π MCP & integrations | 1,524 | MCP servers/clients, third-party integrations, browser automation |
| π§ͺ Testing & QA | 1,506 | Test generation, E2E, coverage, QA process, browser testing |
| π¨ Frontend & design | 1,359 | UI/UX, design systems, CSS, React, accessibility |
| βοΈ Writing & content | 1,345 | Copywriting, editing, humanizing, translation, narrative |
| π Docs & knowledge mgmt | 1,158 | Documentation, ADRs, technical writing, note systems |
| πΌ Business & finance | 1,152 | Finance, pricing, legal, sales, HR, exec advisory |
| π¬ Media & creative | 778 | Video, image, slides/PPTX, audio, game dev, generative art |
| π Data science & ML | 722 | ML/DL, training, fine-tuning, evaluation, statistics |
| π Product & PM | 473 | PRDs, roadmaps, user stories, discovery, analytics |
| π± Mobile & platform | 398 | iOS/Android/Flutter, desktop, extensions, BI platforms |
| β‘ Productivity & workflow | 353 | Planning, automation, meetings, scaffolds, coaching |
| π RAG, memory & knowledge | 333 | RAG, vector search, embeddings, agent memory, knowledge graphs |
| π§© Skill authoring meta | 11 | Creating, testing, auditing and improving skills themselves |
Full index with descriptions and navigation: INDEX.md
the-mega-skill-library/
βββ INDEX.md master index β start here
βββ CATALOG.tsv one row per asset β grep/awk this
βββ master-skill/SKILL.md router skill β install this one, it finds the rest
βββ library/<category>/<slug>/SKILL.md 18,681 Agent Skills (+ bundled resources)
βββ agents/<category>/<slug>.md 2,888 subagent definitions
βββ commands/<category>/<slug>.md 2,773 slash commands
βββ prompts/<category>/<slug>.md 10,074 system prompts / corpora
βββ rules/<category>/<slug>.md 1,386 CLAUDE.md / AGENTS.md / .cursorrules
βββ reference/<owner~repo>/... 543 awesome-lists, cheatsheets, handbooks
βββ _index/<category>.md per-category browsable tables
βββ _logs/ harvest run logs
The raw upstream git clones used to build this (
_sources/, ~9 GB, 186 repos) aren't part of this repository β only the curated, deduplicated output above is published. See Refreshing below.
This is a curated aggregation, not original work. Every skill, agent, command, prompt and rules file was written by its original author β full attribution for all 237 source repositories lives in CREDITS.md, linked directly to each one's GitHub page.
Every single harvested file also carries source_repo, source_path and
source_url in its own frontmatter, plus a human-readable Source footer
linking upstream β open any file to see exactly where it came from. Duplicates
across repos were collapsed by content hash, keeping the most authoritative
copy (official vendor repos first, then by star count); collapsed copies are
listed in each file's "Also appears in" footer, so nothing is silently lost.
This repo's own layer β the categorization, indexing, catalog, router skill, and documentation β is MIT licensed. The 35,802 harvested files themselves come from 186 different repositories (plus 51 more kept as reference material) under their own original licenses β check the upstream repo (linked in every file's footer, and in CREDITS.md) before redistributing or using an individual file commercially.
The curated buckets here (library/, agents/, commands/, prompts/,
rules/, reference/, _index/, CATALOG.tsv) were generated once from
local clones of all 186 source repos. To regenerate: re-clone each repo listed
in CATALOG.tsv / _index/by-source-repo.md, then re-run a categorize β
dedup-by-hash β catalog pipeline against them. See CONTRIBUTING.md
if you want to help formalize that pipeline into a script others can run.
Found a skill-bearing repo that isn't here yet? Found a broken path or a
duplicate that should've been collapsed? PRs and issues are very welcome β
see CONTRIBUTING.md for the frontmatter convention and
how to add an entry to CATALOG.tsv.
SKILL.md bundle (+ optional scripts/templates) an agent loads on demand for one task./slash-command shortcut that expands to a fixed prompt or workflow.CLAUDE.md / AGENTS.md / .cursorrules style always-on instruction file.No. The router (master-skill/) is the only thing that ever loads, and it's
one small file. It searches CATALOG.tsv on demand and reads just the one or
two matching files per task β the other 35,800 assets sit untouched on disk.
The curation layer is MIT. Individual harvested files retain their original upstream license β check the source before commercial use of a specific file. See Provenance & licensing.
Either it hasn't been discovered yet, or it didn't fit the harvest criteria
(skill-shaped content: SKILL.md, subagent defs, slash commands, system
prompts, or rules files). Open a PR β see Contributing.
If this saved you time, star the repo β it's the main way other people find it. Issues, PRs, and "hey you should add X" suggestions are all welcome.
Made by @thedixitjain Β· MIT Licensed
4 commits
Python
75.9%
JavaScript
7.5%
Shell
5.4%
TypeScript
3.3%
HTML
2.3%
Go Template
1.6%
35,802 agent skills, subagents, slash commands, prompts and rules files curated from 186 GitHub repositories β browsable, greppable, installable.
See the code35,802 ready-to-use skills, subagents, slash commands, prompts and rules β harvested, deduplicated and organized from 186 repositories across the Claude Code / Cursor / Codex / Copilot skill ecosystem, so you never have to write one from scratch again.
Browse the index Β· Search the catalog Β· Quick start Β· Credits Β· Contribute
AI coding agents (Claude Code, Cursor, Codex, Copilotβ¦) got a superpower this year: Skills β small, self-contained instruction packs that teach an agent how to do one thing well (write a spec, review a PR, design an API, run a security audit, format a resume). The problem: they're scattered across hundreds of individual GitHub repos, half-documented, and impossible to search in one place.
This repo fixes that. Every skill-bearing repo in the ecosystem, pulled together, deduplicated, categorized, and made greppable in one file. Whether you're a student who just discovered agent skills exist, or a staff engineer wiring up a fleet of subagents in production β the same catalog works for you.
| Kind | Count | Where | What it is |
|---|---|---|---|
| π§© Agent Skills | 18,681 | library/ | Full SKILL.md bundles (+ scripts, templates, references) |
| π€ Subagents | 2,888 | agents/ | Specialized agent role definitions |
| β‘ Slash Commands | 2,773 | commands/ | /command shortcuts |
| π¬ Prompts | 10,074 | prompts/ | System prompts, prompt corpora, leaked prompts |
| π Rules | 1,386 | rules/ | CLAUDE.md / AGENTS.md / .cursorrules exemplars |
| π Reference | 543 | reference/ | Awesome-lists, cheatsheets, handbooks (verbatim) |
git clone https://github.com/thedixitjain/the-mega-skill-library.git
Pick the path that matches how much you already know:
Copy the router skill. It searches the other 35,801 assets for you β nothing else loads into context or slows your agent down.
cp -r the-mega-skill-library/master-skill ~/.claude/skills/master-skill-library
Now just talk to Claude Code normally. When a task matches something in the library, it finds the right skill and uses it automatically. That's it β you're done.
Claude Code reads every installed skill's frontmatter at startup, so only install what you'll actually use β dozens, not thousands.
cp -r the-mega-skill-library/library/engineering-core/tdd-guide ~/.claude/skills/
Project-only instead of global? Copy into <project>/.claude/skills/ instead
of ~/.claude/skills/.
grep -i "pdf" CATALOG.tsv
awk -F'\t' '$2=="security-and-compliance" && $1=="skill"' CATALOG.tsv
grep -i "code review" CATALOG.tsv | awk -F'\t' '$1=="agent"' | cut -f3,6
CATALOG.tsv is one row per asset: kind Β· category Β· name Β· path Β· source_repo Β· description.
Most files here are plain Markdown with YAML frontmatter β portable by design.
library/ bundles follow the Agent Skills spec (works with any harness that
supports it); rules/ files map directly to .cursorrules / AGENTS.md /
custom instructions; prompts/ and commands/ are copy-paste text you can
drop into any system-prompt or slash-command config.
22 categories, auto-sorted by asset count. Click through for the full browsable table of every skill/agent/command/prompt in that category.
| Category | Assets | Covers |
|---|---|---|
| π¦ General purpose | 5,575 | Uncategorized / cross-cutting |
| π€ AI agents & harness | 4,583 | Agent design, orchestration, subagents, context engineering |
| π¬ Prompt engineering | 2,868 | Prompt techniques, system prompts, optimizers, persona design |
| βοΈ Engineering core | 2,805 | Refactoring, debugging, review, architecture, git, TDD |
| π¬ Research & academic | 2,483 | Literature review, papers, journals, grants, regulatory science |
| βοΈ DevOps & infra | 1,605 | Docker, K8s, Terraform, CI/CD, cloud, observability, SRE |
| π Security & compliance | 1,604 | AppSec, pentest, OSINT, threat modeling, SOC2/ISO/GDPR |
| π£ Marketing & growth | 1,584 | SEO/AEO, ads, CRO, social, campaigns, growth loops |
| ποΈ Backend & data | 1,583 | APIs, databases, schemas, queues, caching, pipelines |
| π MCP & integrations | 1,524 | MCP servers/clients, third-party integrations, browser automation |
| π§ͺ Testing & QA | 1,506 | Test generation, E2E, coverage, QA process, browser testing |
| π¨ Frontend & design | 1,359 | UI/UX, design systems, CSS, React, accessibility |
| βοΈ Writing & content | 1,345 | Copywriting, editing, humanizing, translation, narrative |
| π Docs & knowledge mgmt | 1,158 | Documentation, ADRs, technical writing, note systems |
| πΌ Business & finance | 1,152 | Finance, pricing, legal, sales, HR, exec advisory |
| π¬ Media & creative | 778 | Video, image, slides/PPTX, audio, game dev, generative art |
| π Data science & ML | 722 | ML/DL, training, fine-tuning, evaluation, statistics |
| π Product & PM | 473 | PRDs, roadmaps, user stories, discovery, analytics |
| π± Mobile & platform | 398 | iOS/Android/Flutter, desktop, extensions, BI platforms |
| β‘ Productivity & workflow | 353 | Planning, automation, meetings, scaffolds, coaching |
| π RAG, memory & knowledge | 333 | RAG, vector search, embeddings, agent memory, knowledge graphs |
| π§© Skill authoring meta | 11 | Creating, testing, auditing and improving skills themselves |
Full index with descriptions and navigation: INDEX.md
the-mega-skill-library/
βββ INDEX.md master index β start here
βββ CATALOG.tsv one row per asset β grep/awk this
βββ master-skill/SKILL.md router skill β install this one, it finds the rest
βββ library/<category>/<slug>/SKILL.md 18,681 Agent Skills (+ bundled resources)
βββ agents/<category>/<slug>.md 2,888 subagent definitions
βββ commands/<category>/<slug>.md 2,773 slash commands
βββ prompts/<category>/<slug>.md 10,074 system prompts / corpora
βββ rules/<category>/<slug>.md 1,386 CLAUDE.md / AGENTS.md / .cursorrules
βββ reference/<owner~repo>/... 543 awesome-lists, cheatsheets, handbooks
βββ _index/<category>.md per-category browsable tables
βββ _logs/ harvest run logs
The raw upstream git clones used to build this (
_sources/, ~9 GB, 186 repos) aren't part of this repository β only the curated, deduplicated output above is published. See Refreshing below.
This is a curated aggregation, not original work. Every skill, agent, command, prompt and rules file was written by its original author β full attribution for all 237 source repositories lives in CREDITS.md, linked directly to each one's GitHub page.
Every single harvested file also carries source_repo, source_path and
source_url in its own frontmatter, plus a human-readable Source footer
linking upstream β open any file to see exactly where it came from. Duplicates
across repos were collapsed by content hash, keeping the most authoritative
copy (official vendor repos first, then by star count); collapsed copies are
listed in each file's "Also appears in" footer, so nothing is silently lost.
This repo's own layer β the categorization, indexing, catalog, router skill, and documentation β is MIT licensed. The 35,802 harvested files themselves come from 186 different repositories (plus 51 more kept as reference material) under their own original licenses β check the upstream repo (linked in every file's footer, and in CREDITS.md) before redistributing or using an individual file commercially.
The curated buckets here (library/, agents/, commands/, prompts/,
rules/, reference/, _index/, CATALOG.tsv) were generated once from
local clones of all 186 source repos. To regenerate: re-clone each repo listed
in CATALOG.tsv / _index/by-source-repo.md, then re-run a categorize β
dedup-by-hash β catalog pipeline against them. See CONTRIBUTING.md
if you want to help formalize that pipeline into a script others can run.
Found a skill-bearing repo that isn't here yet? Found a broken path or a
duplicate that should've been collapsed? PRs and issues are very welcome β
see CONTRIBUTING.md for the frontmatter convention and
how to add an entry to CATALOG.tsv.
SKILL.md bundle (+ optional scripts/templates) an agent loads on demand for one task./slash-command shortcut that expands to a fixed prompt or workflow.CLAUDE.md / AGENTS.md / .cursorrules style always-on instruction file.No. The router (master-skill/) is the only thing that ever loads, and it's
one small file. It searches CATALOG.tsv on demand and reads just the one or
two matching files per task β the other 35,800 assets sit untouched on disk.
The curation layer is MIT. Individual harvested files retain their original upstream license β check the source before commercial use of a specific file. See Provenance & licensing.
Either it hasn't been discovered yet, or it didn't fit the harvest criteria
(skill-shaped content: SKILL.md, subagent defs, slash commands, system
prompts, or rules files). Open a PR β see Contributing.
If this saved you time, star the repo β it's the main way other people find it. Issues, PRs, and "hey you should add X" suggestions are all welcome.
Made by @thedixitjain Β· MIT Licensed
4 commits
Python
75.9%
JavaScript
7.5%
Shell
5.4%
TypeScript
3.3%
HTML
2.3%
Go Template
1.6%