--- title: Top AI/ML & Agent Tooling Skills slug: ai-ml-agent-tooling type: primary-category canonical: https://top-agent-skills.com/top-skills-for-ai-ml-agent-tooling skills: 22 rubric: 1.0 updated: 2026-08-19 --- # Top AI/ML & Agent Tooling Skills **antigravity-awesome-skills** is the top-ranked ai agent tooling Skill at 65/100: 1,234+ skills in one bundle. Single install across AI coding assistants. Build skills, MCP servers, and agent infra. Claude API SDK, skill-creator, mcp-builder. 22 Skills indexed, ranked by a published composite score (https://top-agent-skills.com/about/methodology). Content last changed 2026-08-19. ## The ranking ### 1. antigravity-awesome-skills — score 65/100 (community) 1,234+ skills in one bundle. Single install across AI coding assistants. - Best for: Teams that want a comprehensive starting set rather than picking 30 individually. - Not ideal for: Lean setups — you will install many skills you never use. - Publisher: Antigravity · MIT - Verified on: claude-code, cursor, codex, gemini-cli, antigravity ```bash npx antigravity-awesome-skills --claude ``` Full entry: https://top-agent-skills.com/skill/antigravity-awesome-skills (https://top-agent-skills.com/skill/antigravity-awesome-skills.md) ### 2. create-payment-credential — score 57/100 (verified-org) Stripe's answer to agentic checkout: one-time payment credentials so an agent can buy without ever seeing the underlying card. - Best for: Letting an agent complete a real purchase on a merchant site, with the human approving each transaction and the real credentials never exposed. - Not ideal for: Anything resembling a sandbox. This spends real money on live merchant sites — treat it as a production capability with a human in the loop, never as a dev-time convenience. - Publisher: Stripe · MIT - Verified on: claude-code ```bash npx skills add stripe/link-cli --skill create-payment-credential ``` Full entry: https://top-agent-skills.com/skill/stripe-link-checkout (https://top-agent-skills.com/skill/stripe-link-checkout.md) ### 3. dbhub — score 56/100 (verified-org) One skill across Postgres, MySQL, SQL Server, MariaDB and SQLite — and its actual job is forcing explore-before-query on real schemas. - Best for: Querying an unfamiliar production-shaped schema, where the agent guessing table and column names is the failure you are trying to prevent. - Not ideal for: Setups without MCP. It requires the DBHub MCP server plus a DSN, and it can execute SQL — so scope the connection to a read-only role unless writes are the point. - Publisher: Bytebase · MIT - Verified on: claude-code ```bash npx @bytebase/dbhub@latest --transport http --port 8080 --dsn "postgres://user:password@localhost:5432/dbname" ``` Full entry: https://top-agent-skills.com/skill/dbhub (https://top-agent-skills.com/skill/dbhub.md) ### 4. composio — score 55/100 (verified-org) 1,000+ SaaS integrations as one skill. OAuth lifecycle, scoped credentials, structured actions. - Best for: Any agent that needs to act on real SaaS systems — Slack, Notion, GitHub, Stripe, Salesforce. - Not ideal for: Read-only research where MCPs would be lighter. - Publisher: Composio · Apache-2.0 - Verified on: claude-code, claude-ai, claude-api, cursor, codex, gemini-cli, antigravity ```bash npx skills add composiohq/skills ``` Full entry: https://top-agent-skills.com/skill/composio (https://top-agent-skills.com/skill/composio.md) ### 5. qdrant-advisor — score 55/100 (verified-org) A self-updating vector-database skill: it fetches current Qdrant guidance at run time instead of shipping a snapshot that rots. - Best for: Diagnosing a Qdrant deployment that is slow, memory-hungry, or returning poor results — the failure modes where stale advice is worse than none. - Not ideal for: Air-gapped environments. It needs network access to skills.qdrant.tech to load current guidance, which is exactly the design that keeps it fresh. - Publisher: Qdrant · Apache-2.0 - Verified on: claude-code ```bash npx skills add qdrant/skills/meta/qdrant-advisor ``` Full entry: https://top-agent-skills.com/skill/qdrant-advisor (https://top-agent-skills.com/skill/qdrant-advisor.md) ### 6. weaviate — score 52/100 (verified-org) The vector-database skill that ships executable scripts rather than prose — the agent runs the query instead of describing it. Full entry: https://top-agent-skills.com/skill/weaviate-skill (https://top-agent-skills.com/skill/weaviate-skill.md) ### 7. supabase-skill — score 51/100 (verified-org) Supabase done right. RLS policies, edge functions, realtime, auth — in idiomatic patterns. Full entry: https://top-agent-skills.com/skill/supabase-skill (https://top-agent-skills.com/skill/supabase-skill.md) ### 8. stripe-payments — score 51/100 (verified-org) Stripe done right. Checkout, subscriptions, webhooks, idempotency, metered billing. Full entry: https://top-agent-skills.com/skill/stripe-payments (https://top-agent-skills.com/skill/stripe-payments.md) ### 9. shopify-storefront — score 51/100 (verified-org) Shopify Storefront API: cart, checkout, metafields, customer accounts, Hydrogen patterns. Full entry: https://top-agent-skills.com/skill/shopify-storefront (https://top-agent-skills.com/skill/shopify-storefront.md) ### 10. setup-cowork — score 50/100 (official-anthropic) Bootstrap the right plugin stack for your role in one guided pass. Full entry: https://top-agent-skills.com/skill/setup-cowork (https://top-agent-skills.com/skill/setup-cowork.md) ### 11. notion-api-skill — score 49/100 (verified-org) Notion API: page/database CRUD, block hierarchies, rich text, schema sync. Full entry: https://top-agent-skills.com/skill/notion-api-skill (https://top-agent-skills.com/skill/notion-api-skill.md) ### 12. mcp-builder — score 48/100 (official-anthropic) Build new MCP servers. The official scaffold + design walkthrough. Full entry: https://top-agent-skills.com/skill/mcp-builder (https://top-agent-skills.com/skill/mcp-builder.md) ### 13. skill-creator — score 48/100 (official-anthropic) Build new skills interactively. Output is a clean SKILL.md ready to publish. Full entry: https://top-agent-skills.com/skill/skill-creator (https://top-agent-skills.com/skill/skill-creator.md) ### 14. claude-api — score 48/100 (official-anthropic) Always-current Claude API reference and SDK docs. Bundled with Claude Code. Full entry: https://top-agent-skills.com/skill/claude-api (https://top-agent-skills.com/skill/claude-api.md) ### 15. init — score 48/100 (official-anthropic) Bootstrap a CLAUDE.md so future sessions start informed. Full entry: https://top-agent-skills.com/skill/init-skill (https://top-agent-skills.com/skill/init-skill.md) ### 16. update-config — score 48/100 (official-anthropic) Edit settings.json safely. Hooks, perms, env, behaviors. Full entry: https://top-agent-skills.com/skill/update-config (https://top-agent-skills.com/skill/update-config.md) ### 17. keybindings-help — score 48/100 (official-anthropic) Rebind keys and add chords without reading the keymap docs. Full entry: https://top-agent-skills.com/skill/keybindings-help (https://top-agent-skills.com/skill/keybindings-help.md) ### 18. fewer-permission-prompts — score 48/100 (official-anthropic) Cuts the permission-prompt friction by learning your safe-tool patterns. Full entry: https://top-agent-skills.com/skill/fewer-permission-prompts (https://top-agent-skills.com/skill/fewer-permission-prompts.md) ### 19. alirezarezvani-skills — score 47/100 (community) 232+ skills across engineering, marketing, product, compliance, advisory. Full entry: https://top-agent-skills.com/skill/alirezarezvani-skills (https://top-agent-skills.com/skill/alirezarezvani-skills.md) ### 20. cowork-plugin-customizer — score 45/100 (official-anthropic) Tailor a plugin to your org's stack in one pass. Full entry: https://top-agent-skills.com/skill/cowork-plugin-customizer (https://top-agent-skills.com/skill/cowork-plugin-customizer.md) ### 21. create-cowork-plugin — score 45/100 (official-anthropic) Scaffold a new plugin from scratch with guided prompts. Full entry: https://top-agent-skills.com/skill/create-cowork-plugin (https://top-agent-skills.com/skill/create-cowork-plugin.md) ### 22. loki-mode — score 38/100 (community) Multi-agent orchestration. Spawn parallel subagents and merge their outputs reliably. Full entry: https://top-agent-skills.com/skill/loki-mode (https://top-agent-skills.com/skill/loki-mode.md) ## About this category The best agent skills for AI/ML and agent tooling in 2026 are Anthropic's official skill-creator, mcp-builder, and claude-api Skills — the canonical patterns for scaffolding new SKILL.md files, building MCP servers in FastMCP or the TypeScript SDK, and integrating the Claude API with prompt caching and the Agent SDK. AI/ML & agent tooling Skills are the meta-layer — Skills that help you build other Skills, MCP servers, agent infrastructure, and Claude API applications. The category centers on Anthropic's official skill-creator, mcp-builder, and claude-api Skills, which encode the canonical patterns for each. Common workflows include scaffolding a new Skill from a single description, running evaluations to test a Skill's triggering accuracy, optimizing a Skill's description for better LLM dispatch, building a new MCP server in Python (FastMCP) or TypeScript (MCP SDK), wiring prompt caching into a Claude API application, migrating Claude API code between model versions, and authoring agent personas with custom system prompts. Several Skills target the upcoming Claude Agent SDK for custom agent runtimes. AI engineers, builders shipping AI products, MCP server authors, and indie developers writing custom Skills use these heavily. Composite scoring is dominated by provenance (Anthropic-authored Skills sit at the top), recency (these Skills age fast as models and SDKs evolve), and the quality of bundled examples — Skills that ship working code samples vastly outperform pure-documentation Skills. ## Common questions **What does skill-creator do?** It scaffolds a new Skill from a description, generates the SKILL.md with proper frontmatter, sets up the directory structure, and includes any helper scripts. It also runs evals to measure how reliably the Skill triggers. **How do I build an MCP server?** Use the mcp-builder Skill — it walks you through FastMCP (Python) or the MCP TypeScript SDK, sets up the tool definitions, handles auth, and tests the server against Claude Code. **When should I migrate Claude API code to a new model?** The claude-api Skill handles migration between Claude 4.5 → 4.6 → 4.7 and replaces references to retired models. Run it after any major Anthropic announcement. **Do these Skills include prompt caching guidance?** Yes — claude-api enforces prompt-cache adoption in new applications, since the 5-minute TTL meaningfully changes how you structure conversations. **Is the Claude Agent SDK supported?** Several Skills target the Agent SDK for custom runtime work — subagents, headless mode, custom tools — but the surface is moving quickly. Check the lastUpdated field. --- _Top Agent Skills. Machine-readable twin of https://top-agent-skills.com/top-skills-for-ai-ml-agent-tooling._ _JSON: https://top-agent-skills.com/top-skills-for-ai-ml-agent-tooling.json · whole catalog: https://top-agent-skills.com/skills.json (CC BY 4.0)_