Install
$ agentstack add mcp-roxx0x-headcount ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
headcount
1,000 skills is a landfill. This is a payroll of 44.
Every other list races to the biggest number. This one is a hiring decision: 44 agent skills, each with a job description, each link verified by hand. Clone one with ./hire.sh and it lands in .claude/skills/.
./hire.sh obra/superpowers DietrichGebert/ponytail
Compatible with Claude Code · Cursor · Codex · Gemini CLI · OpenCode — anything that reads SKILL.md.
Built on the format Anthropic publishes for Agent Skills.
Every link verified 2026-07-23. Star counts are real, pulled from the GitHub API on that date — not copied from a screenshot. Run ./scripts/check-links.sh to re-verify the whole roster yourself.
Why a payroll instead of a directory
The big lists have stopped being lists. The largest one has 1,077 open issues and hasn't been touched since May 22, 2026 — two months of dead links nobody swept. It also opens with a vendor banner and a UTM-tagged section for the vendor's own product. You came for an index and walked into a funnel.
A thousand entries is not curation. It is a landfill with a table of contents.
So this repo inverts it. Hundreds looked at, forty-four kept. Each one gets the thing a category label never gives you — when to hire it. No vendor sections. No affiliate links. No UTM parameters. No sponsor tier. If an entry is here, it is because it earns a seat.
The roster (44 hires, 11 departments)
⏸ = solid, but no commits in the last 30 days. Still worth hiring, just not actively moving.
Engineering
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | superpowers | 259,621 | MIT | Agentic skills framework and development methodology | You want one opinionated loadout instead of assembling forty | | ECC | 232,324 | MIT | Harness optimization: skills, instincts, memory, security | Your agent works but you can't tell why it sometimes doesn't | | skills | 183,124 | MIT | Matt Pocock's .agents directory, TypeScript-first | You write TS and want taste baked in, not explained | | agent-skills | 79,937 | MIT | Production-grade engineering skills | You ship to real users and "it compiles" isn't the bar | | ponytail | 88,037 | MIT | Makes the agent think like the laziest senior in the room | Your agent pulls a library in to format a date | | get-shit-done | 64,783 | MIT | Meta-prompting and context engineering system | You keep re-explaining the same context every session | | planning-with-files | 25,644 | MIT | File-based planning with crash-proof recovery | Long runs die halfway and take the plan with them | | skills | 163,527 | — | Anthropic's own Agent Skills repo | You want the reference format straight from the source |
Design
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | ui-ux-pro-max-skill | 109,113 | MIT | Design intelligence for UI/UX across platforms | Everything your agent builds looks like a Bootstrap demo | | taste-skill | 66,553 | MIT | Stops generic, boring output; aesthetic judgment | The output is correct and completely forgettable | | huashu-design | 21,838 | MIT | HTML-native prototypes, slides, animation, MP4 export | You need a clickable prototype, not a description of one | | hallmark ⏸ | 15,821 | MIT | Anti-AI-slop design skill for Claude Code, Cursor, Codex | You can smell the AI in your own landing page |
Marketing & Growth
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | marketingskills | 41,284 | MIT | CRO, copywriting, SEO, analytics, growth engineering | You built the thing and nobody arrived | | claude-seo | 12,102 | MIT | 25 sub-skills, 18 sub-agents: technical SEO, schema, local | You need SEO done as a pipeline, not as advice |
Memory & Context
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | caveman | 92,176 | MIT | Cuts ~65% of tokens by talking like a caveman | Your bill grew faster than your codebase | | claude-mem | 88,294 | Apache-2.0 | Persistent context across sessions, compressed and injected | Every new session starts with you re-introducing yourself | | rtk | 72,692 | Apache-2.0 | CLI proxy cutting 60–90% of tokens on common dev commands | git diff alone is eating your context window | | mem0 | 61,509 | Apache-2.0 | Universal memory layer; ships as an official Claude connector | You want a real memory store, not a markdown file | | context7 | 59,617 | MIT | Up-to-date library docs piped to the model | Your agent confidently uses an API that was removed in v3 |
Codebase Intelligence
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | graphify | 94,102 | Apache-2.0 | Turns a codebase, SQL schemas, configs, PDFs into a queryable graph | Grep stopped scaling three refactors ago | | Understand-Anything | 75,718 | MIT | Code into interactive, explorable knowledge graphs | You inherited a repo and nobody left a map | | qmd ⏸ | 28,172 | MIT | Local BM25 + vector + LLM re-rank search over your docs | Your knowledge base is local and must stay local |
Research & Science
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | scientific-agent-skills | 31,518 | MIT | 148 skills across biology, chemistry, drug discovery | The work is science, not CRUD | | ARIS | 13,737 | MIT | Markdown-only skills for autonomous ML research loops | You want experiments running while you sleep | | AI-Research-SKILLs | 11,006 | MIT | Research and engineering skills library, model-agnostic | You need the agent to read papers, not just code |
Security
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | Anthropic-Cybersecurity-Skills | 26,385 | Apache-2.0 | 817 skills mapped to MITRE ATT&CK, NIST CSF 2.0 | Your review needs to name a framework, not a vibe |
Ops & Orchestration
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | agency-agents | 135,984 | MIT | A full agency: eng, design, growth, PM, QA, legal | One agent is a person; you need a company | | gstack | 123,784 | MIT | Garry Tan's exact setup: CEO, Designer, Eng Manager, QA | You want a proven loadout, not a blank config | | ruflo | 65,604 | MIT | Meta-harness for multi-agent swarms and autonomous flows | Coordination is now the hard part, not the work | | agents | 38,162 | MIT | Multi-harness plugin marketplace across six agent CLIs | Your team is not all on the same tool | | oh-my-claudecode | 37,990 | MIT | Teams-first multi-agent orchestration, zero learning curve | You want orchestration without reading a manual |
Method & Harness
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | learn-claude-code | 71,996 | MIT | A nano agent harness built from fundamentals | You want to understand the loop, not just run it | | claude-code-best-practice | 63,331 | MIT | Vibe coding → agentic engineering, staged | You're past prompting and don't know what's next | | BMAD-METHOD | 50,994 | see repo | 12+ agents: brainstorm → PRD → architecture → shipped | You need a process, not a smarter model | | harness-engineering | 2,212 | CC-BY-4.0 | Anthology, field guide, agent context bundle | You reference docs from CLAUDE.md and want good ones | | sample-specship | 208 | MIT | recon → plan → build → validate → ship, adversarial gates | Small repo, real idea: the builder can't grade itself |
Ecosystem & Tooling
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | cc-switch | 120,307 | MIT | Desktop switcher across Claude Code, Codex, OpenCode, Grok | You run more than one agent CLI and hate the juggling | | awesome-mcp-servers | 91,212 | MIT | The de-facto MCP server directory | Skills cover thinking; MCP covers touching real tools | | obsidian-skills ⏸ | 43,037 | MIT | Obsidian CLI, Markdown, Bases, JSON Canvas | Your second brain should be writable by the agent | | knowledge-work-plugins | 22,967 | Apache-2.0 | Anthropic's plugins for knowledge work in Claude Cowork | The job function isn't engineering | | skills | 10,855 | Apache-2.0 | Drive the Hugging Face ecosystem from the terminal | Models, datasets and Spaces are part of your loop | | skills ⏸ | 5,639 | MIT | Anthony Fu's curated personal collection | You trust taste over volume |
HR (skills that make skills)
| hire | ⭐ | license | what it does | hire when | |---|---|---|---|---| | SkillSeekers | 14,531 | MIT | Turns docs sites, repos and PDFs into skills, with conflict detection | The skill you need doesn't exist yet | | baoyu-skills | 24,033 | MIT | Skills to write, illustrate and publish articles | The output is content, and it has to read like a human |
Hiring
hire.sh clones a skill into .claude/skills/ and strips its .git so your repo stays clean.
# one hire
./hire.sh obra/superpowers
# several at once
./hire.sh DietrichGebert/ponytail Leonxlnx/taste-skill JuliusBrussee/caveman
# somewhere else
SKILLS_DIR=~/.claude/skills ./hire.sh addyosmani/agent-skills
Verifying
Don't take the numbers on faith — that is the whole complaint about the other lists.
./scripts/check-links.sh
Walks every GitHub link in this README and reports anything that no longer resolves. data/skills.json holds the same roster machine-readable, with the star counts and the date they were pulled.
Contributing
One rule: a pull request must argue for a seat, not for an entry. Say what the skill does and when you'd hire it. See [CONTRIBUTING.md](CONTRIBUTING.md).
MIT. Built to be forked.
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: Roxx0x
- Source: Roxx0x/headcount
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.