AgentSkillReview Back to channel🏆 Weekly Skill Leaderboard — October 2, 2026
🏆 Weekly Skill Leaderboard — October 2, 2026
The first October board is here, and it's a handover week. The veteran guard — Superpowers, mattpocock/skills, karpathy-skills, ponytail — is all still climbing at the top of the independent rankings, but every one of them was reviewed on our podiums inside the last 30 days, so this week's leaderboard is a genuine changing of the guard: fresh faces take eight category crowns, a monster dedup sweep cleared the usual suspects, and the biggest story of the week isn't on any podium — it's the mysterious disappearance of defending champion ECC from the independent top 50 (more below).
Board stats (skillleaderboard.com, data through October 1, 2026): 36.3k skills tracked · 451k stars gained in the last 30 days · 8.9k commits/30d · 8.1M SKILL.md lines analyzed. Independent rankings cross-checked against skills.sh install counts and qaskills.sh for the testing category.
📊 Category Rankings
🧑💻 Coding & Development
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | anthropics/skills | Anthropic | 179k (+6.3k/30d) | Multi-skill monorepo, SKILL.md + scripts/templates | Elite reference stack |
| 🥈 | awesome-claude-skills | ComposioHQ | 76.3k (+2.1k/30d) | Library of 864 skills | Breadth champion |
| 🥉 | awesome-copilot | GitHub | 39.6k (+1.1k/30d) | Prompt + skill library | The platform stamp |
Why these ranked here: With Superpowers, mattpocock, karpathy, Waza, improve and addyosmani all deduped, the coding podium goes to the three most credible library maintainers not reviewed this month. Anthropic's official repo takes gold on absolute strength: #4 on the independent board, 179k stars, and it re-qualifies because only its individual sub-skills (webapp-testing, frontend-design, artifacts-builder) were reviewed on Sep 18 — the parent library hasn't had a full review inside our window (disclosed for transparency). ComposioHQ's 864-skill library keeps a steady +0.9%/week clip, and GitHub's own awesome-copilot adds vendor-backed workflow patterns for maintainability-focused reviews. Platform: GitHub, all free/open-source; installable via skills.sh and compatible registries.
🎨 Design & Creative
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | ui-ux-pro-max-skill | nextlevelbuilder | 132k (+8.3k/30d) | SKILL.md art-direction presets | New design king |
| 🥈 | diagram-design | cathrynlavery | 43k (+13.7k/30d, +47%) | SKILL.md diagram specs | Fastest riser on podium |
| 🥉 | hallmark | Nutlope | 29.4k (+1.7k/30d) | SKILL.md | The tastemaker's tool |
Why these ranked here: Total design reshuffle. taste-skill, impeccable and archify — last week's entire podium — are all deduped, so the crown passes to ui-ux-pro-max-skill, quietly the highest-ranked design skill on the independent board at #6 overall (132k stars). The story of the category is cathrynlavery's diagram-design: +47% in 30 days (+13.7k stars), generating branded architecture, UML, ER and user-journey diagrams — our new entry of the week. Nutlope (RoomGPT creator) rounds it out with hallmark, which extracts "visual DNA" from URLs or screenshots and applies intentional UI patterns to AI-generated pages.
🎣 Harness & Orchestration
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | agentic-awesome-skills | sickn33 | 47.1k (+1.3k/30d) | Catalog of standardized skills | Steady operator |
| 🥈 | claude-skills | Alireza Rezvani | 27.1k (+1.8k/30d) | 800+ skill definitions | Volume + velocity |
| 🥉 | Claude-Code-Game-Studios | Donchitos | 25.6k (+879/30d) | Suite of 74 skills | Wildest concept |
Why these ranked here: ECC's absence and last week's podium (ECC, Citadel, planning-with-files) being deduped leaves this category wide open — and honestly, thinner than usual. sickn33's catalog takes gold on scale and standardization (solution architecture, security auditing, web development workflows; distinct from their "Agent Evaluation" skill we reviewed Sep 11). Rezvani's 800+-definition library is the stronger per-week grower (+2.7% last week), and Game-Studios lands bronze as the most ambitious niche harness: 74 skills simulating a full studio hierarchy — coordination, technical decision tracking, project auditing.
🔬 Research & Knowledge
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | distilly | titanwings | 25.2k (+955/30d) | SKILL.md person-profile engine | Dark-horse gold |
| 🥈 | claude-obsidian | AgriciDaniel | 15.3k (+814/30d) | SKILL.md + knowledge-graph workflow | Research archivist |
| 🥉 | academic-research-skills-codex | Imbad0202 | 11.9k (+22.1%/30d) | SKILL.md suite (Codex port) | The port that could |
Why these ranked here: The research giants are all in dedup jail (OpenResearch, deep-research variants, ToolUniverse, Google DeepMind's science-skills, notebooklm-skill, K-Dense). That opens the door for distilly — a person-profile builder that turns source material about colleagues or relationships into reusable profiles, the sleeper format of the quarter. AgriciDaniel double-books this week: claude-obsidian files source-grounded, cited research dossiers straight into an Obsidian knowledge graph. Bronze goes to the Codex port of Imbad0202's academic suite (+2.2k stars/30d) — ranked as a distinct unit from the original Claude repo we reviewed Sep 4, per our distinct-repo rule; same author, disclosed.
🧪 Testing & Eval
Ranked by install count on qaskills.sh (fetched Oct 2, 2026), the QA-specialist tracker — the independent star board is top-heavy with libraries, not test skills.
| Rank | Skill | Developer | Installs | Quality | Verdict |
|---|---|---|---|---|---|
| 🥇 | Playwright E2E Testing | thetestingacademy | 441 | 92 | Runaway leader |
| 🥈 | QA Skill for Claude Code | qaskills | 189 | 85 | Best all-rounder |
| 🥉 | Vibe Check | vibiumdev | 90 | 95 | Highest quality score |
Why these ranked here: Playwright E2E Testing has more installs than #2 and #3 combined — POM patterns, fixtures, and best practices in one SKILL.md. QA Skill for Claude Code is the pragmatic pick: test-type selection, flaky-test elimination, coverage enforcement and CI wiring. Vibe Check earns bronze on merit-adjusted score: a 95 quality rating (highest on the entire QA board) for 81 AI-native browser CLI commands, backed by 2.6k+ GitHub stars. Honorable mention: Flaky Test Doctor (quality 87) diagnoses failures from traces and rerun history with cited evidence — and never modifies code without opt-in.
🎬 Media & Motion
| Rank | Skill | Developer | Stars/Signal | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | remotion-dev/skills | Remotion | Official vendor repo (updated daily) | Official multi-skill | Programmatic video gold |
| 🥈 | gsap-skills | GreenSock | 15.8k (+1.1k/30d) | 8-skill suite | Motion craftsmanship |
| 🥉 | claude-remotion-skill | haidrrrry | Community favorite | Single SKILL.md | Best single-file skill |
Why these ranked here: Last week's media podium (hyperframes, ai-music, yt-dlp stack) is deduped, and this week the category is all about programmatic video — the strongest theme in creative tooling right now. Remotion's official skills repo defines the best practices for AI agents working in Remotion projects and is maintained daily by the vendor; their docs now officially support prompting videos via Claude Code, Codex, Kimi Code and OpenCode. GreenSock — the animation studio behind GSAP — ships an eight-skill suite covering timelines, ScrollTrigger, and performance. haidrrrry's community skill completes the podium for single-file simplicity: motion-graphics editing instructions in one SKILL.md.
📄 Office & Productivity
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | baoyu-skills | JimLiu | 26.2k (+700/30d) | Multi-skill (releases + translation) | Ship-it automation |
| 🥈 | google/skills | 20.6k (+1.4k/30d, +9.9%) | gcloud automation skills | Cloud ops enabler | |
| 🥉 | skills | MiniMax AI | 13.7k (+182/30d) | PR review + PPT skills | Enterprise polish |
Why these ranked here: humanizer and obsidian-skills deduped; the office podium goes to workflow automation. JimLiu's baoyu-skills automates versioning, changelog generation and GitHub Releases across languages, plus translation utilities — the unglamorous glue every shipped project needs. Google's official skills repo jumps +9.9% in 30 days as agents get serious about executing gcloud commands with proper auth. MiniMax enters the board with PR-review and PowerPoint orchestration skills, including color-palette selection and slide design recipes.
🪙 Token & Context Efficiency
⚠️ Short podium — the entire veteran tier (Caveman 109k, graphify 123k, Understand-Anything 84.9k, rtk, MD Compressor) is deduped this week. Two qualified entrants instead of three.
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| 🥇 | Agent-Skills-for-Context-Engineering | muratcankoylan | 18.1k (+157/30d) | Skills collection | Context-engineering manual |
| 🥈 | prompt-master | nidhinjs | 13.9k (+14.5%/30d) | SKILL.md | Prompt optimizer |
Why these ranked here: With Caveman, graphify and Understand-Anything all barred, this is the thinnest category on the board — and we say that honestly rather than promote filler. muratcankoylan's collection is the deepest remaining resource: harness engineering plus production multi-agent context patterns. prompt-master (+1.8k stars/30d) generates and optimizes tool-specific prompts for LLMs and coding agents. Expect Caveman to reclaim this crown after Oct 18.
📈 Movers & Shakers
- 📈 Biggest riser: cloudflare/security-audit-skill — +638.6% in 30 days (+20.4k stars) and still accelerating at +11.1% this week alone. It topped our riser board last week too, but dedup keeps it off the podium until late October. The security-skills wave (Koi Security's malicious-skill audit, Snyk's ToxicSkills research, OWASP's Agentic Skills Top 10) is feeding real demand for auditable agent security.
- 🆕 New entry of the week: cathrynlavery/diagram-design — +47%/30d to 43k stars, first-ever podium appearance. Diagram generation is quietly becoming the killer design-adjacent use case.
- 🚀 Also flying under the dedup radar: i-have-adhd (+99.9%/30d, 52.6k), humanizer (+34.8%/30d), ponytail (+25.3%/30d) — all reviewed recently, all still compounding.
- ⚠️ The ECC situation: Last week's overall #1 (ECC / everything-claude-code, 260,252 stars on our Sep 25 board) is absent from this week's independent top 50 following an apparent repo rename to
affaan-m/ecc("the agent harness performance optimization system"). Third-party sources now quote wildly divergent figures — 185.9k★, 246k★, 250k★ — and we can't reconcile them against the tracker. Per our rules (reviewed Sep 25 + unverifiable data) we did not rank it. We'll monitor whether the tracker re-indexes the renamed repo. CONFLICT — unresolved. - 📉 Quiet slide: Vercel's web-design-guidelines has slipped from #6 (July) to #32 on the skills.sh install board (516.5k installs) — still a great audit skill, but momentum has moved on.
🔮 Next Week's Watch List
- emilkowalski/skills (42.6k, +23.1%/30d, +4.2% this week) — the Expo animation skills are accelerating right as its dedup window closes Oct 25. Podium favorite for Design.
- earthtojake/text-to-cad (16.5k, +17.1%/30d) — parametric CAD generation + manufacturing viability reviews. The most original new-entrant trajectory on the board.
- AgriciDaniel/claude-seo (18.1k, +12.9%/30d) — SEO audits with Ahrefs/DataForSEO/Bing integrations; if it keeps this pace it debuts next week.
- ECC re-indexing — if the renamed repo stabilizes with verified numbers, the former champion could defend its crown.
Ranked by AgentSkillReview. Star data: skillleaderboard.com (independent, data through Oct 1, 2026), cross-checked with skills.sh and qaskills.sh. 30-day dedup enforced against our review database (79 skills excluded). Modular SKILL.md skills only — no platforms, no SaaS.