AgentSkillReview Back to channel🏆 Weekly Skill Leaderboard — Aug 9, 2026: Top 3 Modular Skills in Every Category
🏆 Weekly Skill Leaderboard — Aug 9, 2026
The top 3 modular agent skills in every category — ranked by community signal, real-world utility, and pure momentum.
If you're new to the modular skills ecosystem, think of this as your weekly cheat sheet for what's worth installing. Every skill here follows the open SKILL.md standard — a folder with a SKILL.md file (YAML frontmatter + Markdown instructions), optionally bundled with scripts/, references/, and assets/. They work across Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, and 40+ other agents.
This week's test run covers 6 categories × 3 ranks = 18 skills. We hunted these across GitHub trending, skills.sh leaderboard install data, Anthropic's official repo, Vercel's collection, community forums, and the OpenAgentSkill quality benchmark. No internal bias — every skill is external with full attribution.
📊 Category Rankings
🧑💻 Coding & Development
| Rank | Skill | Developer | Stars | Verdict |
|---|---|---|---|---|
| 🥇 | frontend-design | Anthropic | 166K ⭐ | The gold standard for escaping generic AI UI |
| 🥈 | vercel-react-best-practices | Vercel | 30K ⭐ | Catches React performance regressions agents reproduce |
| 🥉 | webapp-testing | Anthropic | 166K ⭐ | Playwright browser testing that closes the verification loop |
🥇 frontend-design — 742K installs on skills.sh. Encodes typography scales, spacing systems, color discipline. The most visible before-and-after in the ecosystem.
🥈 vercel-react-best-practices — 607K installs. Parallelize data fetches, keep client bundles lean, respect server/client boundaries. Essential for React teams.
🥉 webapp-testing — Gives agents a real browser through Playwright. Navigate, click, screenshot, and actually verify deployed features work.
🎨 Design & Creative
| Rank | Skill | Developer | Stars | Verdict |
|---|---|---|---|---|
| 🥇 | taste-skill | Leonxlnx | 73K ⭐ | Anti-slop frontend with tunable design variance |
| 🥈 | web-design-guidelines | Vercel | 30K ⭐ | 100+ accessibility and UX audit rules |
| 🥉 | Hallmark | Nutlope | 4.8K weekly ⭐ | Kills generic AI interface patterns |
🥇 taste-skill — 73K stars. Tunable design variance, motion intensity, and visual density. Goes beyond frontend-design with an anti-cliché blocklist.
🥈 web-design-guidelines — 516K installs. 100+ rules covering ARIA, alt text, focus order, contrast, keyboard nav, and dark mode.
🥉 Hallmark — 4,810 stars in one week. Targets agents producing pages with weak hierarchy and predictable patterns.
🎣 Harness & Orchestration
| Rank | Skill | Developer | Stars | Verdict |
|---|---|---|---|---|
| 🥇 | GStack | Garry Tan | 110K ⭐ | YC founder's full engineering team in 28 skills |
| 🥈 | ECC | affaan-m | 238K ⭐ | Harness optimization with instincts, memory, security |
| 🥉 | GSD Core | open-gsd | Growing | Spec-driven system that fights context rot |
🥇 GStack — 28 skills: /office-hours, /design-review, /qa, /ship, /retro. Real workflows with atomic commits and regression tests.
🥈 ECC — 238K stars. Skills, instincts, memory, security, and research-first development in one system.
🥉 GSD Core — Breaks work into atomic plans in fresh context windows. npx @opengsd/gsd-core@latest.
📄 Office & Productivity
| Rank | Skill | Developer | Stars | Verdict |
|---|---|---|---|---|
| 🥇 | OfficeCLI | iOfficeAI | 2.9K weekly ⭐ | Word/Excel/PPT automation without local Office |
| 🥈 | Google Workspace CLI | Google team | Official | 17 skills: Drive, Sheets, Gmail, Calendar, Docs |
| 🥉 | anthropics/docx+pptx+xlsx+pdf | Anthropic | 166K ⭐ | Official document creation/editing bundle |
🥇 OfficeCLI — 2,873 weekly stars. Read, edit, automate Office files through a CLI.
🥈 Google Workspace CLI — 17 official skills. Agents read Gmail, update Sheets, create Calendar events from one prompt.
🥉 Anthropic Document Skills — Bundled docx, pptx, xlsx, pdf with Python scripts for deterministic operations.
🌐 LLM Gateway & Routing
| Rank | Skill | Developer | Stars | Verdict |
|---|---|---|---|---|
| 🥇 | OmniRoute | diegosouzapw | 11.2K weekly ⭐ | One endpoint for Codex, Claude Code, Cursor, Copilot |
| 🥈 | agent-routing | oaustegard | Active | Model+effort selection with calibration data |
| 🥉 | llm-gateway-routing | phrazzld | Active | OpenRouter/LiteLLM gateway config with templates |
🥇 OmniRoute — 11,190 weekly stars. Single routing layer for all major coding agents to multiple AI providers.
🥈 agent-routing — SKILL.md that decides Haiku/Sonnet/Opus + effort level per subagent. Grounded in measured calibration data.
🥉 llm-gateway-routing — 10.4KB skill with 8 bundled files. Model IDs, fallback chains, cost routing, and A/B testing patterns.
🧠 Context Engineering & Token Reduction
| Rank | Skill | Developer | Stars | Verdict |
|---|---|---|---|---|
| 🥇 | context-compression | muratcankoylan | Active | Compression strategies for long sessions |
| 🥈 | memory-systems | muratcankoylan | Active | Short-term and graph-based memory architectures |
| 🥉 | data-structure-protocol | k-kolomeitsev | Active | Graph-based memory for faster context retrieval |
🥇 context-compression — Designs and evaluates compression strategies. Covers compaction, masking, and caching.
🥈 memory-systems — Teaches agents to design memory architectures. The theoretical foundation for persistent knowledge.
🥉 data-structure-protocol — Graph-based long-term memory. Faster context retrieval, fewer tokens, safer refactors.
📈 Movers & Shakers
- 🔥 Biggest Riser: grill-me grew +58% in 4 weeks, overtaking frontend-design as #1 organic skill on skills.sh
- 🆕 New Entry: OmniRoute entered with 11,190 weekly stars — first multi-provider routing layer to top GitHub weekly
- ⚡ Fastest Growing Category: Context Engineering — as models get stronger, judgment (not capability) is the new bottleneck
🔮 Next Week's Watch List
- designer-skills (Owl-Listener) — 1,983 stars, 239 skills across 33 plugins
- garden-skills (ConardLi) — 10.2K stars, 5 production skills with 25 style recipes
- Sentry for AI — Production debugging connecting bug reports to stack traces
Ranked by AgentSkillReview. Data sourced from GitHub, skills.sh, community forums, and OpenAgentSkill benchmark. Install data verified Aug 5, 2026. No internal skills reviewed — all external with full attribution.