Agent Skills — Coleção de Skills Cross-Platform
JPeetz/agent-skills
Skills cross-platform curadas semanalmente para agentes IA
Ver no GitHubRepositório curado de alta qualidade com skills prontas para produção compatíveis com Claude Code, Cursor, Copilot e outros agentes IA. Organiza 100+ skills em 14 categorias (DevOps, Segurança, Backend, QA, Frontend, Dados, etc.) com documentação, testes e validação rigorosa. Instalação via GitHub CLI ou manual, com suporte a versionamento.
Instalação
npx convez-ai JPeetz/agent-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Expandir capacidades de Claude Code, Cursor e Copilot com skills prontas
- Automatizar CI/CD, IaC, testes e workflows DevOps
- Gerar código frontend, auditorias de acessibilidade e design-to-code
- Análise de dados, design de schemas e dbt pipelines
- Segurança de agentes, compliance GDPR e automação de releases
Veredito

Repositório robusto e bem-organizado que eleva o padrão de qualidade de skills para agentes. Instalação flexível, documentação clara e avaliação rigorosa justificam a posição como referência do ecossistema. Ideal para quem busca skills production-ready em vez de protótipos.
README
A tradução para o português deste README ainda está na fila.
Agent Skills — The Highest-Quality Collection of Cross-Platform Agent Skills
The most trusted, most cited, most useful skill source in the Agent Skills ecosystem.
Every skill is cross-platform compatible: Claude Code, Codex, Cursor, OpenClaw, Gemini CLI, GitHub Copilot, Hermes — any SKILL.md-compatible agent.
Structure
Skills are organized by category under skills/:
skills/
├── devops-infrastructure/ CI/CD, IaC, Kubernetes, observability, SRE
├── security/ Agent & supply-chain security
├── api-backend/ API design, GraphQL, HAR reverse-engineering
├── qa-testing/ Playwright E2E, browser automation
├── frontend/ a11y audits, design-to-code
├── data/ Analysis, dbt, database schema design
├── documentation-content/ Docs, document processing, X articles
├── skill-development/ Skill lifecycle, mining, authoring
├── business-strategy/ App discovery, Linear PM
├── development/ Scaffolding, cost routing, engineering habits
├── marketing/ Ad campaigns, image generation
├── compliance-legal/ Privacy & GDPR
├── git-release/ Git workflow & release automation
└── code-quality/ Code review
Install
# Via GitHub CLI (recommended) — gh skill launched April 2026
gh skill install JPeetz/agent-skills
# Install a specific skill (discovered under skills/<category>/<skill>)
gh skill install JPeetz/agent-skills app-discovery-scrutiny --agent claude-code --scope user
gh skill install JPeetz/agent-skills astra-campaign --agent copilot --scope user
# Pin to a version
gh skill install JPeetz/agent-skills app-scaffolding --pin v1.0.0
# Manual install (cross-client, works everywhere) — pick the category path
cp -r skills/<category>/skill-name ~/.agents/skills/
# Examples
cp -r skills/development/app-scaffolding ~/.agents/skills/
cp -r skills/devops-infrastructure/ci-cd-pipeline-generator ~/.agents/skills/
# Platform-specific manual installs
cp -r skills/<category>/skill-name ~/.openclaw/workspace/skills/ # OpenClaw
cp -r skills/<category>/skill-name ~/.claude/skills/ # Claude Code
cp -r skills/<category>/skill-name .codex/skills/ # Codex
cp -r skills/<category>/skill-name .cursor/skills/ # Cursor
Why This Repo
Most skill repositories optimize for volume. We optimize for quality. Every skill here has been:
- Evaluated against a 10-dimension scoring framework
- Improved over existing public alternatives (never shipped unchanged)
- Validated with real test cases and assertions (eval suite under
evals/evals.json) - Documented with usage examples and platform notes (corrections logs added where a skill's history has one)
- Packaged with self-contained scripts, eval suites, and changelogs
Skills Catalog
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| ci-cd-pipeline-generator | DevOps/CI-CD | Production-ready CI/CD pipelines — GitHub Actions, GitLab CI, CircleCI, Jenkins | Universal |
| infrastructure-as-code-guardian | DevOps/Infrastructure | Cross-tool IaC security and management — Terraform, Pulumi, CloudFormation, Ansible, Bicep | Universal |
| kubernetes-operations | DevOps/K8s | Production-grade Kubernetes ops — manifest generation, Helm charts, GitOps, security hardening, failure-mode diagnosis | Universal |
| observability-engineering | DevOps/SRE | OpenTelemetry instrumentation, monitoring, distributed tracing, SLI/SLO management, incident response | Universal |
| production-engineering-workflows | DevOps/SRE | Full SDLC automation: /spec → /plan → /build → /test → /review → /ship | Universal |
| sre-runbooks | DevOps/SRE | Safe-by-default SRE runbooks — incident response, postmortems, on-call handovers, RCA | Universal |
| maintenance | DevOps/Ops | Self-hosted customer-stack maintenance runbook — JEV-first decisions, health probes, weekly backup, JEV-gated updates, rollback, sparse LLM cost discipline | Universal |
| performance-optimization | DevOps/Perf | Frontend, backend, and infrastructure performance optimization | Universal |
| ci-cd-and-automation | DevOps/CI-CD | Quality-gated build/deploy pipelines — nothing ships without passing gates | Universal |
🛡️ Security (4 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| agentic-security-scanner | Security/Agent | OWASP AST10 security scanning — malicious skill detection, prompt injection, data exfiltration, SARIF | Universal |
| supply-chain-security-scanner | Security/DevSecOps | Software supply chain security — SBOM generation, dependency scanning, provenance verification, license compliance | Universal |
| threat-modeling | Security/Threats | STRIDE threat model with risk-rated threats and backlog stories | Universal |
| security-and-hardening | Security/Hardening | OWASP Top 10 hardening — three-tier boundary, dependency audit, LLM safety | Universal |
🔌 API & Backend (8 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| api-design-first | API/Backend | Design-first OpenAPI 3.1 specifications — REST, GraphQL, gRPC with cross-protocol consistency | Universal |
| graphql-api-development | API/Backend | GraphQL API design and optimization — Apollo Federation, DataLoader, security, subscriptions | Universal |
| har-api-reverse-engineering | API/Backend | Reverse-engineer a site's hidden/undocumented API from a captured HAR — derive, replay, verify, reuse (authorized use only) | Universal |
| api-and-interface-design | API/Design | Design APIs, module boundaries, and public interfaces that are hard to misuse | Universal |
| social-har-api-connectivity | API/Social | Connect a social platform's API from a captured HAR — filter, derive, verify, reuse (authorized only) | Universal |
| image-to-image-character-generation | API/Image | Generate consistent character images from Supabase/GDrive refs via kie.ai image-to-image — face anchors, dynamic prompt, anti-repeat | Universal |
| kie-ai-platform | API/Platform | Full kie.ai platform API: image/video/audio/chat gen, async tasks, callbacks, file upload, credits, webhook HMAC verifcation | Universal |
| n8n-api-shape | API/Automation | Instance-agnostic n8n REST API — read/update/verify workflows, strip-extras PUT recipe, banned fields, health check | Universal |
🧪 QA & Testing (4 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| browser-automation | QA/Browser | Playwright browser automation — E2E testing, web scraping, monitoring, form submission, screenshots | Universal |
| browser-testing-with-devtools | QA/Browser | Test in real browsers via Chrome DevTools MCP — DOM, network, a11y | Universal |
| playwright-e2e-testing | QA/Testing | Production-grade Playwright E2E testing — locator strategy, CI/CD, visual regression, component testing, a11y | Universal |
| test-driven-development | QA/TDD | Write a failing test before the code that makes it pass | Universal |
🎨 Frontend (3 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| accessibility-compliance-audit | Frontend/A11y | WCAG 2.2 AA audits — automated scan, manual review, fix-ready code for React/Vue/Angular | Universal |
| frontend-ui-engineering | Frontend/UI | Accessible, responsive, polished production UI — React/Vue/Svelte/HTML | Universal |
| design-to-code | Frontend/Design | AI-powered design-to-code: Figma, Sketch, screenshots → production React/Vue/Svelte/HTML | Universal |
📊 Data (4 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| data-analysis | Data Science | Data analysis, visualization, and reporting for CSV, JSON, Excel, SQL datasets | Universal |
| exploratory-data-analysis | Data Science | Explore unfamiliar datasets, data-quality triage, pipeline validation | Universal |
| dbt-data-transformation | Data/DBT | Production-grade dbt analytics engineering — model development, testing, dbt Mesh governance, semantic layer | Universal |
| database-schema-designer | Database | Production-ready schema design — normalization, indexing, safe migrations, multi-tenant | Universal |
📝 Documentation & Content (5 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| document-processing | Content/Docs | PDF/DOCX/XLSX/PPTX automation — generation, manipulation, conversion, OCR, mail merge | Universal |
| read-x-articles | Content/Web | Read X (Twitter) long-form Articles end-to-end from a shared link via the canonical /i/article/ID URL + render-capable extraction | Universal |
| spec-writer | Documentation/Specs | Turn vague feature requests into structured specs, technical plans, and task breakdowns — inline [ASSUMPTION] markers for every unspecified decision | Universal |
| documentation-and-adrs | Documentation/ADRs | Record architecture decisions and document design rationale | Universal |
| technical-documentation | Documentation | AI-powered technical docs: READMEs, ADRs, API docs, runbooks, knowledge bases | Universal |
🛠️ Skill Development (3 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| skill-lifecycle-foundry | Skill Development | Mine, author, personalize, and generalize agent skills across their full lifecycle — session mining to release checklist | Universal |
| skill-miner | Skill Development | Mine agent session/archive history for repeated skill-worthy workflows | Universal |
| interview-me | Skill Development | One-question-at-a-time interview to resolve underspecified asks | |
| skill-portability-and-packaging | Skill Development | Package and port skills across agent hosts: spec hard rules, install roots, description budgeting, plugin/marketplace manifests, hazard sweep, install verification |
|### 💼 Business & Strategy (3 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| app-discovery-scrutiny | Business/AI | VC-grade app niche evaluation — Build/Pivot/Kill verdict | Universal |
| linear-project-management | Business/PM | Drive Linear issues, projects, and teams from an agent via MCP + GraphQL — discovery-before-create, label taxonomy | Universal |
| enterprise-consulting-methods | Business/Strategy | 15 McKinsey/BCG-grade methodology patterns — SCPR, issue/hypothesis trees, storyline building, decision memos, stakeholder maps, meeting prep, workshop design, native PowerPoint chart production via python-pptx | Universal |
|### ⚙️ Development (11 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| app-scaffolding | Development | 12-section mobile app build blueprint (iOS + Android) | Universal |
| model-routing-cost-optimizer | Development/Cost | Route agent tasks to the cheapest adequate model tier — ROUTINE/MODERATE/COMPLEX classifier + vision guard | Universal |
| anti-over-engineering | Development | Keep AI agents from over-engineering code beyond the user's request | Universal |
| joerg-instruction-execution | Development | Execute shorthand/high-trust instructions by intent + verify output-side | Universal |
| finding-unknowns | Development | Systematic discovery of unknowns when entering unfamiliar code or domain territory | Universal |
| hermes-bot-team-design | Development | Design a team of specialist Hermes bots from a project spec — roles, SOUL.md, MCP wiring, group chat | Universal |
| mattpocock-dev-workflows | Development | Matt Pocock's engineering workflow patterns — spec synthesis, deep-module design, domain modeling, two-axis code review, disciplined bug diagnosis | Universal |
| nodejs-development | Development/Node | Full-stack Node.js development — Fastify patterns, TypeScript type stripping, V8/libuv internals, OAuth 2.0/2.1, graceful shutdown, error handling, streams, testing | Universal |
| repo-recon | Development | Systematic codebase recon before touching unfamiliar code — map, learn commands, trace a flow, match conventions, write it down | Universal |
| telegram-topics-pipeline | Development | Run multiple bot sessions via Telegram forum topics with isolated context | Universal |
| safe-refactor | Development | Change code structure without changing behavior in small verified steps — green baseline, characterization tests, blast-radius tracking | Universal |
📣 Marketing (5 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| astra-campaign | Marketing | Full-service ad campaign generator (X, TikTok, Instagram) | Universal |
| betterlife-image-generation | Marketing/Image | BetterLife daily social images via kie.ai grok-imagine — Mira character anchors, anti-repetition | Universal |
| local-video-production-m5 | Marketing/Video | Local video production on the 16GB Apple-Silicon M5: LTX/Wan GGUF-MLX, memory discipline, character consistency, OpenMontage | Hermes + M5 |
| openmontage | Video | Produce/edit video via OpenMontage pipelines (avatar-spokesperson, talking-head, cinematic) with director layers + approval gates | Hermes |
| agent-n-video-pipeline | Video | Fully-local "Agent N presents a message": Kokoro TTS → LTX-2.3 MLX A2V talking render + lip sync → karaoke → QA → X caption (Herald port, zero cloud) | Hermes + M5 |
⚖️ Compliance & Legal (3 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| privacy-compliance | Compliance/Legal | Global privacy compliance: GDPR, CCPA/CPRA, HIPAA, EU AI Act, LGPD, cross-border transfers | Universal |
| fsi-compliance-checker | Compliance/FSI | Map changes to PCI-DSS / SOX / FSI control IDs | Universal |
| data-privacy-compliance | Compliance/Legal | GDPR/CCPA data privacy — PII, DSRs, consent, breach response | Universal |
🌿 Git & Release (3 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| git-workflow-automation | Git/Release | Conventional Commits, changelogs, PR descriptions, semantic versioning, branch management | Universal |
| release-management | Git/Release | Version bumps, changelogs, tags, release notes, semver | Universal |
| git-workflow-and-versioning | Git/Workflow | Clean commits, branching, conflict resolution, messy-tree splitting | Universal |
|### ✅ Code Quality (5 skills)
| Skill | Domain | Description | Platforms |
|---|---|---|---|
| code-review | Code Quality | AI-powered code review — security, quality, style, architecture, test coverage, performance | Universal |
| code-review-and-quality | Code Quality | Multi-axis review of any code change before merge — own, agent, or human | Universal |
| clean-pro-review-gates | Code Quality | 5-gate pre-merge review framework — code quality (15 LLM failure modes), tests, docs, security (OWASP/CWE/LLM Top 10), CI/CD/container/IaC hardening | Universal |
| debugging-and-error-recovery | Code Quality | Systematic debugging — failing tests, broken builds, regressions | Universal |
| code-simplification | Code Quality | Refactor for clarity without changing behavior | Universal |
FAQ
What are Agent Skills? Agent Skills are a lightweight, open format (agentskills.io) for giving AI agents specialized capabilities. Each skill is a folder with a SKILL.md file containing instructions, plus optional scripts, references, and test suites.
Which platforms support Agent Skills? Claude Code, Claude CLI, Claude Projects, Claude API, OpenAI Codex, Gemini CLI, Cursor, OpenClaw, GitHub Copilot, Hermes Agent, OpenCode, Amp, Junie, and any SKILL.md-compatible agent.
Why are skills nested under categories?
gh skill and the agentskills.io specification discover skills at skills/*/SKILL.md and skills/{scope}/*/SKILL.md — category folders map to the {scope} position, so gh skill install, gh skill publish, and the agentskills.io CLI all resolve nested skills natively. Categories keep a 40-skill library browsable.
How do I install a skill?
Via GitHub CLI: gh skill install JPeetz/agent-skills [skill-name]. Or manually: copy the skill directory to your agent's skills folder. Common paths: ~/.agents/skills/ (cross-client), ~/.claude/skills/ (Claude Code), .codex/skills/ (Codex), .cursor/skills/ (Cursor), ~/.openclaw/workspace/skills/ (OpenClaw).
How often are new skills added? New skills ship every Tuesday and Thursday. Each skill is researched, scored, improved, and validated before publication.
What domains do these skills cover? Business strategy (app discovery, market analysis), development (app scaffolding, API design-first, GraphQL API development), Frontend/Design (design-to-code, accessibility compliance), DevOps/SRE (CI/CD pipeline generation, infrastructure-as-code guardian, production engineering workflows, SRE runbooks, Kubernetes operations, observability engineering), Git workflow automation (commits, PRs, changelogs, releases), code quality (AI-powered code review covering security, architecture, performance), data (data analysis and visualization, dbt analytics engineering, database schema design), compliance and privacy (GDPR, CCPA, HIPAA, EU AI Act, cross-border transfers), QA and testing (Playwright E2E testing, browser automation), security (supply chain scanner, agentic security scanner, SBOM, dependency scanning, provenance verification), content and documents (PDF/DOCX/XLSX/PPTX processing, technical documentation), marketing (ad campaign generation).
How can I submit a skill? Open an issue or PR. See CONTRIBUTING.md. Skills are reviewed against our quality framework before merging.
Are skills tested? Every skill includes an eval suite (evals/evals.json) with test cases, should-trigger/not-trigger scenarios, and verifiable assertions. Skills are validated against the agentskills.io specification.
Repository Standards
- SKILL.md validation (frontmatter, body limits, description quality)
- Eval suite (min 5 test cases per skill)
- Self-contained scripts (PEP 723 inline dependencies)
- Platform notes when behavior differs across agents
- Changelog per skill
- MIT license
(Note: a corrections log is added per skill only where the skill's real development history produced documented failures — it is not a blanket requirement for every listing.)
Managed by Skill Foundry
This repository is maintained autonomously by Skill Foundry — an AI department of AgentForge. Skills are discovered, evaluated, improved, and published on a twice-weekly cadence. Quality beats volume. Every run leaves the repo better than before.
💛 Support the Project
Agent Skills is free and open source. If it's saving you time and powering your workflow, consider supporting development:
ETH: 0xdfa8ac0f37d1129af72d0c4c6c0dff22e7a816b7
SOL: DogfWxst6B9gkoWquG1eGr1q1ExQeXbx2K1zhg9oEDhB
GitHub Sponsors: Sponsor on GitHub
Every contribution helps keep this project moving. Thank you 🙏
🤝 Contributing
PRs are welcome — skills, improvements, and bug reports. See CONTRIBUTING.md for the submission and review process.
- Bug fixes / skill improvements → open a PR directly
- New skills / feature requests → open an issue first to discuss scope and fit
- Security issues → see SECURITY.md for responsible disclosure
Available via gh skill install JPeetz/agent-skills | Part of the AgentForge Ecosystem | agentskills.io