An agentic software-development methodology: composable skills that push a coding agent through spec, plan, TDD and review instead of straight into code.
Development Skills
Build software, MCP servers, and web apps.
Tencent's skill-evolution harness: it rewrites a whole skill folder — SKILL.md, scripts and references together — and lands every decision as a real Git issue, PR and wiki entry you can review.
Take an OpenSearch search application from requirements to a running cluster — BM25, dense and sparse vectors, hybrid retrieval, agentic search and RAG, with relevance evaluation built in.
Auth0's official agent skill: a router that detects your framework and intent, then loads the right Auth0 guidance for login, MFA, Organizations, tenant audits, debugging or provider migration.
Redis' own guidance for FT.CREATE schema design, FT.SEARCH / FT.AGGREGATE / FT.HYBRID, HNSW vector similarity and RAG retrieval pipelines.
Supabase's official skill covering Database, Auth, Edge Functions, Realtime, Storage, Vectors, Cron and Queues — with a hard rule to verify against the live changelog before writing code.
GreenSock's official ScrollTrigger skill — scroll-linked animation, pinning, scrub and trigger positioning, with the exact start/end syntax agents get wrong.
Microsoft's official skill that turns an agent into an Azure solution architect — 10 design principles, 6 architecture styles, 44 design patterns and a Well-Architected review workflow.
Trail of Bits' security review skill for PRs, commits, and diffs: risk-first analysis with git history, blast radius, and honest coverage limits.
Generates and validates parametric STEP-first CAD parts and assemblies from natural language.
Deno's official foundational skill — project setup, package choices, configuration, and CLI commands done the current way.
Build Gradio web UIs and ML demos in Python with the current API — components, layouts, events, and chat interfaces.
Cloudflare's official platform skill — Workers, Pages, storage, AI, networking, security, and IaC in one reference.
Never claim work is complete, fixed, or passing without running the check and reading the output first.
Enforce a real red-green-refactor loop when implementing features or fixes, so behaviour is specified before it is written.
A disciplined debugging process — reproduce, isolate, hypothesise, and verify — used before proposing any fix.
Explore intent, requirements, and design with the user before any code gets written — the antidote to building the wrong thing quickly.
Elastic's official skill for monitoring LLM and agent workloads — token cost, latency, response quality and workflow orchestration, queried with ES|QL.
MongoDB's official data-modelling skill — embed-vs-reference decisions, relationship patterns and the anti-patterns that quietly wreck performance and cost.
Deploy, configure and troubleshoot the NVIDIA RAG Blueprint from your agent — Docker Compose, Helm or library, with feature toggles for VLM, guardrails and agentic RAG.
NVIDIA's own guidance for moving pandas workloads onto GPU DataFrames with cuDF and dask-cuDF — when it pays off, and how to keep results identical.
A decision board for contested calls: feed your agent papers and half-formed ideas, and it files them into a traceable sources → ideas → output structure you can defend, with a visual canvas over the markdown.
67 game-development skills across ten engines — Godot, Unity, Unreal, Bevy, Phaser, PixiJS, three.js, LÖVE, pygame, Roblox — behind a router that detects your engine and loads only what fits.
A meta-skill that spots the moment your agent has just earned a hard-won golden path and writes it down as a reusable skill, so the next session starts already knowing the route.
Provision and deploy to Aiven for OpenSearch from an agent — managed clusters on AWS, GCP, Azure, DigitalOcean or UpCloud, with the Aiven-generated password read back after provisioning.
Investigate OpenTelemetry traces stored in OpenSearch — slow and error spans, reconstructed trace trees, service maps, plus agent invocation and token-usage analysis.
Query and analyse logs held in OpenSearch using PPL and Query DSL — error-pattern discovery, error-rate tracking and anomaly detection, driven from natural language.
Answer "who changed what" from Datadog Audit Trail — configuration-change forensics, API key compromise, cost-spike root cause, SOC 2/PCI evidence and AI activity audits.
Install, instrument and analyse Datadog APM from an agent — Kubernetes and Linux Single-Step Instrumentation, service remapping, trace search and service maps.
Search Datadog logs from an agent and keep the bill under control — query syntax, exclusion filters, log-based metrics, archives and PII scrubbing.
Create and audit Datadog monitors from an agent via the pup CLI — with alerting rules that stop the usual flapping, unscoped and unowned monitors.
ClickHouse's official agent skill: 31 rules for primary keys, data types, JOINs, batching and partitioning that an agent must check and cite before advising on a schema or query.
Neon's official agent skill for serverless Postgres — connection strings and drivers, pooled vs direct connections, branching, autoscaling, scale-to-zero, read replicas and logical replication.
Netlify's official agent skill for shipping a project with the Netlify CLI — auth, site linking, preview and production deploys, and reading live function logs when a deploy misbehaves.
Stripe's official agent skill for choosing the right payments primitive — Checkout vs PaymentIntents, Connect, Billing, Tax — and for handling API keys and webhooks safely.
Upstash's official skill for the Workflow TypeScript SDK - define serverless workflow endpoints, run durable steps, and trigger and manage runs from the client.
Elastic's official skill for writing and running ES|QL - the piped Elasticsearch query language - covering schema discovery, query patterns, error handling and result formats.
Browserbase's official skill for driving a real browser from an agent via the browse CLI - navigate, snapshot, fill forms and click, locally or in a cloud session.
Clerk's official skill for advanced Next.js auth: middleware strategy, protecting Server Actions, caching-safe auth() usage, session tokens and manual JWT verification.
Prisma's official skill for migrating an ORM project from v6 to v7 - the prisma-client generator, driver adapters, prisma.config.ts, ESM entrypoints and every removed API.
Audit UI code against Vercel's Web Interface Guidelines — accessibility, interaction, performance and layout rules — reported as terse file:line findings.
Vercel Engineering's React and Next.js performance playbook as an agent skill — 70 rules across 8 categories, ordered by impact, for writing and refactoring fast components.
Integrate Firecrawl's /scrape endpoint into product code — turn a known URL into markdown, HTML, links, metadata or screenshots, with sensible defaults baked in.
Sentry's Django performance review skill — hunts N+1 queries, unbounded querysets, missing indexes and write loops, and reports only findings it can prove.
Analyse a repository's tech stack and generate a recommended Claude Code settings.json permission allowlist for its read-only commands.
Cypress' official skill for writing, updating and repairing E2E and component tests — and it triggers on "write tests" even when nobody says Cypress.
Angular's official skill for generating idiomatic modern Angular — signals, resource and httpResource, forms, DI, routing, SSR, ARIA and CLI scaffolding.
Resend's official skill for the whole email lifecycle — transactional sends, inbound webhooks, templates, broadcasts, domains and delivery events.
Qdrant's meta-skill: instead of shipping static docs, it loads the current official Qdrant skill tree live from skills.qdrant.tech and diagnoses from that.
MongoDB's official skill for choosing between Atlas Search, Vector Search and Hybrid Search, then building the right indexes and queries for the use case.
The Flutter team's official skill for structuring an app in the recommended UI / Logic / Data layers, with MVVM view-models and the repository pattern.
Firebase's official red-team skill for auditing Firestore and Cloud Storage security rules — hunting privilege escalation, create/update bypasses and resource-exhaustion holes.
Apollo's official skill for wrapping REST APIs into a GraphQL supergraph with @source and @connect directives — no resolvers, with compose, run and test steps enforced.
File-based planning for long-running agent tasks — crash-proof markdown plans that survive /clear and compaction, with per-turn re-injection and a deterministic completion gate.
The .NET team's official skill for scanning C# against ~50 performance anti-patterns across async, memory, strings, collections, LINQ, regex and I/O.
Google's official Android skill for auditing R8 and ProGuard keep rules — finds redundant and overly broad rules that quietly inflate APK size.
Remotion's official skill for scaffolding a programmatic video project in React and iterating on it in the Studio preview instead of re-rendering.
GreenSock's official skill for the GSAP core engine — tweens, eases, staggers, defaults and responsive/reduced-motion animation via matchMedia.
Microsoft's official skill for giving AI agents real OAuth2 identities in Entra — blueprints, agent identities, sponsors, workload identity federation and the auth sidecar.
Google's official skill for calling Gemini on Agent Platform (formerly Vertex AI) with the Gen AI SDK — Python, JS/TS, Go, Java and C#, plus tools, multimodal input and the Live API.
Google's official skill for getting Google Cloud authentication right — human identities, service accounts, Application Default Credentials, and the failure modes agents usually hit.
PostHog's official skill for auditing a live analytics implementation — scripts, network calls, and window state — read-only, via browser automation.
NVIDIA-verified skill that takes an agent from a problem described in prose to a correct LP, MILP, or QP formulation — concepts, not API calls.
Official Expo skill for Expo Router: file-based routes, native stacks, NativeTabs, modals and form sheets, headers, toolbars, and search bars.
Grafana's official PromQL skill: write, validate, and optimise Prometheus queries, and hunt down the label that blew up your cardinality.
Skill: GitHub Actions Security Review
by Sentry
Audit GitHub Actions workflows for the exploitation patterns that leak secrets or allow code execution from a fork.
Skill: Code Simplifier
by Sentry
Clean up code for clarity, consistency, and maintainability without changing what it does.
Skill: Find Bugs in Branch Changes
by Sentry
Audit the changes on your current branch for bugs, security vulnerabilities, and quality problems before you push.
Skill: Engineering Code Review
by Sentry
Review pull requests against a real engineering practice — security, performance, testing, and design, in that order.
Skill: Deno Sandbox
by Deno
Run untrusted or model-generated code in an isolated environment using the @deno/sandbox SDK.
Skill: Deno Frontend with Fresh
by Deno
Build web UIs with Fresh and Preact on Deno — routes, handlers, islands, and Tailwind, using the current framework API.
Skill: Deno Deploy
by Deno
Deploy Deno applications to production on Deno Deploy — CLI workflows, environment config, KV, and custom domains.
Skill: Local Model Selection
by Hugging Face
Choose and run the right local model with llama.cpp and GGUF — quantisation, hardware fit, and local serving.
Skill: LLM Fine-Tuning with TRL
by Hugging Face
Fine-tune language and vision models with TRL or Unsloth — supervised, preference, and reinforcement training methods.
Skill: Transformers.js
by Hugging Face
Run machine learning models directly in the browser or Node with Transformers.js — no inference server required.
Skill: Hugging Face Spaces
by Hugging Face
Deploy and maintain applications on Hugging Face Spaces — SDK choice, GPU hardware, model loading, and debugging.
Skill: Hugging Face CLI
by Hugging Face
Hugging Face's official CLI skill — download, upload, and manage models, datasets, Spaces, and Jobs from the terminal.
Skill: Web Performance Auditing
by Cloudflare
Audit Core Web Vitals with real measurements via Chrome DevTools — then fix what the numbers actually point to.
Skill: Cloudflare Sandbox SDK
by Cloudflare
Build secure code-execution features on Cloudflare — sandboxes for AI-generated code, interpreters, and CI systems.
Skill: Wrangler CLI
by Cloudflare
Cloudflare's CLI reference skill — correct Wrangler commands for deploying and managing Workers and every binding.
Skill: Cloudflare Durable Objects
by Cloudflare
Create and review Cloudflare Durable Objects — single-instance coordination with SQLite storage, alarms, and WebSockets.
Skill: Cloudflare Agents SDK
by Cloudflare
Build stateful AI agents on Cloudflare Workers — durable state, scheduling, RPC, WebSockets, and MCP servers.
Skill: Writing Skills
by Jesse Vincent
Create, edit, and verify agent skills — including how to write a description that actually triggers at the right moment.
Skill: Dispatching Parallel Agents
by Jesse Vincent
Run two or more genuinely independent tasks concurrently, with the briefing and isolation that makes it safe.
Skill: Using Git Worktrees
by Jesse Vincent
Set up an isolated workspace with git worktrees before starting feature work or executing a plan.
Skill: Subagent-Driven Development
by Jesse Vincent
Execute an implementation plan by dispatching independent tasks to subagents, keeping the main context focused.
Skill: Receiving Code Review
by Jesse Vincent
Respond to code review feedback with technical rigour — verify each point instead of agreeing with everything.
Skill: Requesting Code Review
by Jesse Vincent
Prepare and request a code review that checks the work against its original requirements, not just its syntax.
Skill: Executing Implementation Plans
by Jesse Vincent
Work through a written implementation plan in a fresh session, with review checkpoints between steps.
Skill: Writing Implementation Plans
by Jesse Vincent
Convert a specification into a written implementation plan with discrete, verifiable steps, before touching code.
Skill: Vercel Optimize
by Vercel
Vercel's official optimization skill: collects real production metrics first, then returns ranked, citation-checked cost and performance fixes.
Skill: Data Visualization
by Anthropic
Anthropic's official data-visualization skill: pick the right chart for the data, then build it with ready-to-run matplotlib, seaborn, and plotly patterns.
Skill: MCP Builder
by Anthropic
Anthropic's guide for building high-quality MCP servers end to end — research, tool design, implementation in TypeScript or Python, testing, and a 10-question evaluation suite.


