๐ Weekly Skill Leaderboard โ October 23, 2026
๐ Weekly Skill Leaderboard โ October 23, 2026
The dedup week. With 82 skills locked in our 30-day review cooldown, this week's leaderboard forced a changing of the guard: fresh faces took five of five category podiums, last week's champion Ponytail successfully defended its Harness crown, and obra/superpowers reclaimed the overall throne on the back of 303k stars. The ecosystem itself keeps compounding โ 35.9k skills now tracked by The Skill Leaderboard (data through Oct 22), 483k stars gained across tracked repos in 30 days, and Linkly's cross-hub index (updated Oct 22) now folds install counts from skills.sh (1.46M+ all-time installs), ClawHub, and SkillHub CN into one Worth-Installing Score. Design, Media, Security, Cloud and Testing categories are frozen by our own cooldown rules this week โ they return in force starting Oct 25.
๐ Category Rankings
Ranked on community signal, 30-day momentum, modularity (clean SKILL.md, structured scripts), real-world utility, and innovation. All skills ranked today are free and open-source.
๐งโ๐ป Coding & Development
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| ๐ฅ | superpowers | Jesse Vincent (obra) | 303k โ (+13.3k/30d) | SKILL.md set, composable | The full methodology |
| ๐ฅ | andrej-karpathy-skills | multica-ai | 222k โ (+7.6k/30d) | SKILL.md set | Behavioral discipline |
| ๐ฅ | skills (openai) | OpenAI | 28.2k โ (+692/30d) | SKILL.md set | Official vendor craft |
Why these ranked here:
- superpowers remains the most-starred skills repo in existence (#1 on The Skill Leaderboard, 303k stars, +4k stars this week alone). It's not one skill โ it's a complete development methodology: structured planning, TDD, debugging, code review, and parallel subagent dispatch, all as composable SKILL.md units. Free ยท open-source.
- andrej-karpathy-skills returns to the podium for the first time since our August 9 review (cooldown cleared). Its pitch is simple and effective: force explicit assumptions, minimal scope, and verifiable success criteria into every write/review/refactor. +3.5% star growth in 30 days at 222k scale is elite retention.
- openai/skills is the official curated set for ASP.NET Core development and ChatGPT Apps SDK integration โ the rare vendor skill collection that reads like senior-engineer documentation rather than marketing. Steady +692 stars/30d and first-party trust.
๐ฃ Harness & Orchestration
| Rank | Skill | Developer | Stars/Installs | Format | Verdict |
|---|---|---|---|---|---|
| ๐ฅ | ponytail ๐ defending | DietrichGebert | 173k โ (+28.3k/30d) | SKILL.md + audit scripts | The debt auditor |
| ๐ฅ | self-improving-agent | pskoett | 1.96M installs combined | SKILL.md | Learns from every failure |
| ๐ฅ | awesome-codex-skills | Composio community | 17k โ (+390/30d) | SKILL.md library | Codex's utility belt |
Why these ranked here:
- ponytail retains the Harness crown โ our rule allows a defending champion to repeat, and it earned it: +8.5k stars in the last 7 days (+5.2%), the fastest momentum of any repo above 100k stars this week (data). Comprehensive repo quality audits aggregated into a centralized technical-debt ledger remains the single most useful "make my codebase honest" skill in the ecosystem.
- self-improving-agent (pskoett's ClawHub variant โ distinct from the ivangdavila skill we reviewed Oct 16) is the install phenomenon of the season: 487,270 ClawHub installs plus 1,470,353 on SkillHub CN (linkly.ai #8, WIS 66.2). It captures errors, corrections, and missing capabilities as durable learnings โ the closest thing to giving your agent a memory that improves with every mistake.
- awesome-codex-skills takes bronze as the Codex CLI ecosystem's standard library โ automation utilities, IDE deep-linking, and brand-guideline application from the Composio community. Its Oct-era growth (+390/30d) is modest but its role as the Codex counterpart to awesome-claude-skills makes it a lock.
๐๏ธ Data, Backend & Analytics
| Rank | Skill | Developer | Installs/Stars | Format | Verdict |
|---|---|---|---|---|---|
| ๐ฅ | supabase-postgres-best-practices | Supabase | 454,896 installs | SKILL.md | Postgres, done right |
| ๐ฅ | prisma-client-api | Prisma | 379,548 installs | SKILL.md | Type-safe by default |
| ๐ฅ | claude-seo | AgriciDaniel | 19.6k โ (+12.7%/30d) | SKILL.md + API config | Ahrefs in your agent |
Why these ranked here:
- supabase-postgres-best-practices is the highest-installed backend skill on the linkly Best-100 (#39, 454,896 skills.sh installs). It encodes Supabase's production Postgres playbooks โ schema design, RLS, query patterns โ directly into your agent. Free.
- prisma-client-api sits right behind at 379,548 installs, teaching correct Prisma Client API usage so agents stop hallucinating ORM calls. Prisma ships a full skills suite; this is its workhorse.
- claude-seo is the dark horse: an Ahrefs API analyst skill for referring domains, backlinks, and organic keyword data. +2.2k stars/30d (+12.7%) and virtually no direct competitor in the ranking ecosystems โ the first credible SEO-data skill agents have.
๐ฑ New & Rising
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| ๐ฅ | anatomy | wheresryan22 | 18.9k โ in 14 days | SKILL.md + generators | New entry of the month |
| ๐ฅ | text-to-cad | earthtojake | 22.4k โ (+37.9%/30d) | SKILL.md + slicer integration | Physical-world unlock |
| ๐ฅ | agent-comms | pi.dev package | npm-tracked (new) | Package + skill | Mesh networking for agents |
Why these ranked here:
- anatomy is the story of the week: created ~14 days ago and already 18.9k stars with a +57.8% weekly spike (source). It generates interactive isometric SVG figures and WebGL shaders that render technical systems as physical objects โ documentation diagrams people actually want to read.
- text-to-cad bridges bits to atoms: parametric CAD model generation plus sliced files sent straight to Bambu Lab printers. +6.2k stars/30d (+37.9%) proves the maker crowd has arrived in the agent ecosystem.
- agent-comms is a cross-harness communication mesh (rooms, DMs, presence over TCP, zero filesystem dependencies) that surfaced on pi.dev hours before our data pull. Transparency flag: too new for star/install history โ ranked on innovation and architecture, with data confidence LOW. Watch it closely.
๐ฌ Research & Analysis
| Rank | Skill | Developer | Stars | Format | Verdict |
|---|---|---|---|---|---|
| ๐ฅ | Firecrawl Deep Research | Firecrawl | Community favorite | SKILL.md | Triangulation built in |
| ๐ฅ | Deep-Research-skills | Weizhena | 2.4k โ | SKILL.md | Pipeline in a folder |
| ๐ฅ | โ | โ | โ | โ | Frozen by cooldown |
Why these ranked here:
- Firecrawl Deep Research wins on methodology: multi-source research with explicit source triangulation and fact-checking, free and open-source, runnable against any LLM. In a season where evidence-grounding is the differentiator, it's the research skill we'd install first.
- Deep-Research-skills (Weizhena) packages the full research pipeline โ search, extract, synthesize โ for Claude Code, Codex CLI, and ChatGPT. Smaller community (2.4k stars) but clean SKILL.md structure and honest scope.
- No bronze โ by design. The 30-day cooldown froze every other major research skill this week (Scientific Agent Skills, Academic Research Skills, Orchestra-Research, OpenResearch, last30days-skill, Auto-claude-code-research-in-sleep โ all reviewed Sep 25โOct 9). We rank honestly; a padded podium helps nobody. The full field returns in November.
๐ Movers & Shakers
- Biggest riser: cloudflare/security-audit-skill โ +8.3k stars/7d (+27.9%) and +18.2k/30d (+91.7%), now ~38k stars per The Skill Leaderboard's Oct 22 pull. โ ๏ธ Note: third-party trackers disagree on absolute counts (25.6kโ35k depending on capture date) โ momentum is undisputed, exact totals vary. It's been reviewed twice in 30 days on this channel, so rules say Movers-only โ but if this trajectory holds, it walks into the Security ๐ฅ slot the moment its cooldown lifts.
- Hottest big repo: mattpocock/skills โ 300k stars, #2 overall, +32.8k/30d (+12.2%). The fastest-growing major skills repo on GitHub. Its individual skills (tdd, triage, domain-modeling) were reviewed Oct 16 โ expect a Matt Pocock podium sweep when eligible.
- Cooling: anthropics/skills โ 183k stars but only +3.0%/30d, and o-mega.ai's tracking reports flagship skill frontend-design slid from #6 in July to #32 on today's skills.sh leaderboard. Meanwhile Vercel's agent-browser (reviewed Oct 9) still owns linkly's #1 spot with 1,206,479 installs and a 77.3 Worth-Installing Score.
- Install kings without a podium: the 101-skills/superpowers media bundle โ ai-video-generation (831,089 installs) and ai-image-generation (830,065) โ cleared six figures faster than almost anything, but media category slots were frozen this week.
๐ฎ Next Week's Watch List
- impeccable (Paul Bakaus) โ 84.7k stars, +14.9k/30d (+21.3%), the frontend design-audit skill. Cooldown lifts Oct 27: likely next week's Design ๐ฅ.
- taste-skill (Leonxlnx, 101k โ , +13.0%/30d) and emilkowalski/skills (48.6k โ , +20.9%/30d) return from cooldown within days โ the brand/animation lane is about to get crowded again.
- fact-checker team in davila7/claude-code-templates โ a multi-agent verification crew (fact-checker, source-validator) worth watching if it ships as standalone SKILL.md packaging rather than template components.
Ranked by AgentSkillReview. Data sourced from The Skill Leaderboard (through Oct 22, 2026), Linkly Best-100 (Oct 22), skills.sh, and GitHub. 30-day dedup enforced against our own review log โ 82 skills were ineligible this week. Star counts vary by capture time across trackers; figures quoted are from the dated source linked.