94 professional skills + 7 full-featured plugins for AI coding agents — organized by audience so you can find exactly what you need. Now covering MCP, Claude 4.6 features, Claude Agent SDK, skills authoring, and evals.
Works with Claude Code, Claude Cowork, Cursor, Codex, Windsurf, and 35+ other AI agents.
# Cross-platform skills (works with any AI agent)
npx skills add latestaiagents/agent-skills --all
# Claude Code / Cowork plugins (with MCP integrations)
/plugin marketplace add latestaiagents/agent-skills
/plugin install devops-sre@latestaiagents-agent-skillsWorks with Claude Code, Cursor, Codex, Windsurf, Cline, Aider, and 35+ other AI agents.
npx skills add latestaiagents/agent-skills --allFull plugins with MCP tool integrations and slash commands. Works with Claude Code CLI and Claude Cowork desktop app.
# Step 1: Add our marketplace
/plugin marketplace add latestaiagents/agent-skills
# Step 2: Install plugins you need
/plugin install devops-sre@latestaiagents-agent-skills
/plugin install qa-testing@latestaiagents-agent-skills
/plugin install hr-people-ops@latestaiagents-agent-skills| You Are | Skills CLI | Claude Plugin |
|---|---|---|
| Everyone | npx skills add latestaiagents/agent-skills/skills/safety --all |
/plugin install safety@latestaiagents-agent-skills |
| Developer | npx skills add latestaiagents/agent-skills/skills/developer --all |
/plugin install developer-toolkit@latestaiagents-agent-skills |
| DevOps/SRE | npx skills add latestaiagents/agent-skills/skills/mlops --all |
/plugin install devops-sre@latestaiagents-agent-skills |
| RAG Engineer | npx skills add latestaiagents/agent-skills/skills/rag-architect --all |
/plugin install rag-plugin@latestaiagents-agent-skills |
| Security Engineer | npx skills add latestaiagents/agent-skills/skills/security --all |
/plugin install security-guardian@latestaiagents-agent-skills |
| QA/Testing | — | /plugin install qa-testing@latestaiagents-agent-skills |
| HR/People Ops | — | /plugin install hr-people-ops@latestaiagents-agent-skills |
Skills are instructions that teach AI agents how to handle specific tasks. After installation, your AI assistant automatically knows:
- How to resolve merge conflicts systematically (not just "fix it somehow")
- How to safely run database migrations (backup first, check existing data)
- How to design multi-agent systems (patterns that actually work)
- When to ask for confirmation before destructive operations
No special commands needed — skills activate automatically based on what you're doing.
Safety skills that prevent accidental data loss. Recommended for all users.
npx skills add latestaiagents/agent-skills/skills/safety --all| Skill | What It Does |
|---|---|
destructive-operation-guard |
Core safety protocols for all destructive operations |
migration-safety |
Safe database migrations with backup requirements |
database-safety |
Prevent accidental DELETE, DROP, TRUNCATE |
file-operation-safety |
Protection against rm -rf and bulk deletions |
git-safety |
Guard against force push, reset --hard, history loss |
Git workflows, code intelligence, and debugging tools — 19 skills total.
npx skills add latestaiagents/agent-skills/skills/developer --all| Skill | What It Does |
|---|---|
merge-conflict-surgeon |
Step-by-step conflict resolution with context analysis |
commit-message-crafter |
Conventional commits that tell a story |
branch-strategy-advisor |
GitFlow vs trunk-based — choose what fits |
git-history-detective |
Find exactly when and where bugs were introduced |
rebase-safely |
Interactive rebase without losing work |
git-undo-wizard |
Recover from reset, rebase, and force push disasters |
| Skill | What It Does |
|---|---|
codebase-context-builder |
Create CLAUDE.md and optimal context for AI |
ai-code-reviewer |
Systematic review of AI-generated code |
refactor-with-ai |
Safe, incremental refactoring workflows |
test-generation-patterns |
AI-driven test creation that actually works |
debug-with-ai |
Structured debugging: hypothesize → verify → fix |
doc-sync-automation |
Keep docs updated when code changes |
code-explanation-generator |
Clear explanations for complex code |
| Skill | What It Does |
|---|---|
error-pattern-analyzer |
Group errors by root cause, find patterns |
log-forensics |
Extract insights from log files |
stack-trace-decoder |
Parse and explain complex stack traces |
reproduction-builder |
Create minimal reproduction cases |
performance-profiler |
Identify bottlenecks systematically |
dependency-conflict-resolver |
Untangle version conflicts |
Multi-agent systems, LLMOps, and production AI patterns — 20 skills total.
npx skills add latestaiagents/agent-skills/skills/mlops --all| Skill | What It Does |
|---|---|
agent-supervisor-pattern |
Hub-and-spoke orchestration with reasoning transparency |
agent-mesh-architecture |
Peer-to-peer resilient agent networks |
agent-handoff-protocols |
Clean task delegation between agents |
agent-memory-systems |
Working, short-term, and long-term memory patterns |
agent-tool-routing |
Dynamic tool selection and MCP integration |
agent-error-recovery |
Circuit breakers, retry policies, graceful degradation |
agent-cost-budgeting |
Token allocation across agent swarms |
agent-testing-harness |
Unit and integration testing for agents |
langgraph-workflows |
LangGraph 1.0 state machines and graph patterns |
durable-state-patterns |
Persistent agent state across failures/restarts |
human-in-loop-agents |
Approval workflows and agent supervision |
agent-checkpointing |
Recovery, replay, and debugging patterns |
a2a-protocols |
Agent-to-Agent communication and MCP integration |
| Skill | What It Does |
|---|---|
token-cost-analyzer |
Audit usage with 2026 pricing (Claude, GPT-4, etc.) |
prompt-caching-patterns |
Reduce costs by 73%+ with semantic caching |
model-routing-strategy |
Route to cheaper models when appropriate |
llm-rate-limiting |
Token bucket, exponential backoff, circuit breakers |
ai-audit-logging |
EU AI Act compliant audit trails |
prompt-injection-guard |
Input validation, canary tokens, output filtering |
llm-fallback-chains |
Multi-provider failover strategies |
Build production-ready RAG systems with 2026 best practices — 7 skills total.
npx skills add latestaiagents/agent-skills/skills/rag-architect --all| Skill | What It Does |
|---|---|
hybrid-retrieval |
Vector + keyword search with RRF fusion and reranking |
chunking-strategies |
Optimal chunking: semantic, parent-child, late chunking |
graphrag-patterns |
Knowledge graph + RAG for multi-hop reasoning |
agentic-rag |
Agent-driven retrieval with planning and reflection |
corrective-rag |
Self-healing RAG with document grading and fallbacks |
rag-evaluation |
RAGAS metrics, testing, and benchmarking |
production-rag-checklist |
End-to-end production deployment guide |
Comprehensive security skills including OWASP Top 10 and common security practices — 16 skills total.
npx skills add latestaiagents/agent-skills/skills/security --all| Skill | What It Does |
|---|---|
injection-prevention |
SQL, NoSQL, and command injection prevention |
broken-auth-detector |
Authentication and session security |
sensitive-data-protection |
Encryption, data masking, secure storage |
xxe-prevention |
XML External Entity attack prevention |
access-control-audit |
Authorization and IDOR detection |
security-misconfiguration |
Server, cloud, and app hardening |
xss-prevention |
Cross-site scripting prevention |
insecure-deserialization |
Safe object serialization |
dependency-vulnerability |
Package and supply chain security |
logging-monitoring |
Security event logging and alerting |
| Skill | What It Does |
|---|---|
api-security |
REST and GraphQL API security |
secrets-detection |
Find leaked credentials and API keys |
secure-code-review |
Security-focused code review methodology |
csrf-protection |
Cross-site request forgery prevention |
secure-headers |
HTTP security headers (CSP, HSTS, etc.) |
jwt-security |
Secure JWT implementation |
Five fresh suites covering the highest-leverage topics of 2026: the Model Context Protocol, Claude 4.6's new capabilities, the Claude Agent SDK, meta-skills for authoring, and eval engineering.
Build and consume Model Context Protocol servers that Claude Desktop, Claude Code, Cursor, ChatGPT, and the Agent SDK all speak.
npx skills add latestaiagents/agent-skills/skills/mcp-mastery --all| Skill | What It Does |
|---|---|
mcp-server-authoring |
Build production MCP servers in TypeScript or Python |
mcp-client-integration |
Wire MCP into Claude Desktop, Claude Code, Cursor, custom apps |
mcp-tool-design |
Tool schemas that agents actually pick correctly |
mcp-resource-patterns |
Tool vs resource vs prompt; URI templates; subscriptions |
mcp-auth-oauth |
OAuth 2.1 + PKCE + Dynamic Client Registration for remote MCP |
mcp-transport-stdio-http |
stdio vs Streamable HTTP; sessions, resumability, scaling |
mcp-security-sandboxing |
Threat model, sandboxing, rate limits, audit, injection defense |
Get the most out of Claude 4.6's newest capabilities.
npx skills add latestaiagents/agent-skills/skills/claude-4-6-features --all| Skill | What It Does |
|---|---|
long-context-1m |
Use the 1M context window affordably (with caching) |
extended-thinking |
Budget tokens, interleaved thinking, when it helps |
memory-tool |
Cross-session persistent agent memory |
code-execution |
Run Python sandboxed; data analysis, charts, verification |
prompt-caching-ttl |
5m/1h TTL caching; breakpoints; common mistakes |
computer-use |
Browser/desktop automation; safety; when NOT to use |
Build production agents with less plumbing.
npx skills add latestaiagents/agent-skills/skills/claude-agent-sdk --all| Skill | What It Does |
|---|---|
agent-sdk-quickstart |
First agent in TS or Python; tools, hooks, permissions |
managed-agents-api |
/v1/agents + /v1/sessions server-side runtime |
session-lifecycle |
Create, resume, fork, compact; when to start fresh |
sub-agent-delegation |
Parallel sub-agents for research and specialization |
agent-sdk-streaming |
Stream text, thinking, and tool activity to UIs |
Build skills that actually activate and help users.
npx skills add latestaiagents/agent-skills/skills/skills-authoring --all| Skill | What It Does |
|---|---|
skill-frontmatter |
Write YAML that makes skills fire reliably |
skill-activation-patterns |
Design triggers; disambiguate overlapping skills |
progressive-disclosure |
Structure SKILL.md for the model's attention |
skill-testing |
CI-gated validation, lint, and activation tests |
Measure LLM quality rigorously; catch regressions before users do.
npx skills add latestaiagents/agent-skills/skills/evals --all| Skill | What It Does |
|---|---|
llm-as-judge |
Rubrics, pairwise, calibration, bias mitigation |
eval-dataset-design |
Coverage, stratification, label quality, contamination |
regression-evals |
Significance testing, CI gating, baseline management |
golden-set-maintenance |
Curate 20-100 must-not-regress cases |
cost-quality-tradeoff |
Pareto analysis; when a premium model pays off |
These 7 plugins include MCP tool integrations for connecting to external services like Datadog, PagerDuty, Jira, Pinecone, and more.
Note: For complete plugin documentation, see the official Claude Code plugins guide.
| Plugin | Skills | Commands | MCP Integrations |
|---|---|---|---|
| devops-sre | 8 | 6 | Datadog, PagerDuty, AWS, Kubernetes, Prometheus |
| rag-plugin | 7 | 3 | Pinecone, Weaviate, Qdrant, Neo4j, Elasticsearch |
| security-guardian | 10 | 2 | Snyk, Semgrep, Trivy, Vault, GitHub Security |
| agent-plugin | 13 | 2 | LangSmith, Redis, PostgreSQL, OpenAI |
| developer-toolkit | 19 | 2 | GitHub, GitLab, Jira, Linear, Sentry |
| hr-people-ops | 6 | 2 | Greenhouse, Lever, Workday, BambooHR |
| qa-testing | 6 | 3 | TestRail, Xray, BrowserStack, Jira |
# Step 1: Add our marketplace (one-time)
/plugin marketplace add latestaiagents/agent-skills
# Step 2: Install plugins
/plugin install devops-sre@latestaiagents-agent-skills
/plugin install rag-plugin@latestaiagents-agent-skills
/plugin install security-guardian@latestaiagents-agent-skills
/plugin install agent-plugin@latestaiagents-agent-skills
/plugin install developer-toolkit@latestaiagents-agent-skills
/plugin install hr-people-ops@latestaiagents-agent-skills
/plugin install qa-testing@latestaiagents-agent-skillsEach plugin includes:
- Skills: Auto-activated based on context
- Slash Commands: Quick access to common workflows (e.g.,
/incident-response) - MCP Configs: Pre-configured connections to external tools (optional, requires your API keys)
After installing a plugin, configure MCP integrations:
- Open the plugin's
.mcp.jsonfile - Add your API keys for the services you use
- All MCP integrations are marked
optional: true— only enable what you need
Example (DevOps SRE plugin):
{
"mcpServers": {
"datadog": {
"command": "npx",
"args": ["-y", "@datadog/mcp-server"],
"env": { "DD_API_KEY": "your-key", "DD_APP_KEY": "your-key" }
}
}
}Works with Claude Code, Cursor, Codex, Windsurf, Cline, Aider, and 35+ other AI agents.
# Everything
npx skills add latestaiagents/agent-skills --all
# By category
npx skills add latestaiagents/agent-skills/skills/safety --all
npx skills add latestaiagents/agent-skills/skills/developer --all
npx skills add latestaiagents/agent-skills/skills/mlops --all
npx skills add latestaiagents/agent-skills/skills/rag-architect --all
npx skills add latestaiagents/agent-skills/skills/security --all
# Individual suite
npx skills add latestaiagents/agent-skills/skills/developer/git-mastery --all
# Single skill
npx skills add latestaiagents/agent-skills/skills/developer/git-mastery/git-undo-wizard# Step 1: Add the marketplace (one-time)
/plugin marketplace add latestaiagents/agent-skills
# Step 2: Browse available plugins
/plugin marketplace list
# Step 3: Install what you need
/plugin install devops-sre@latestaiagents-agent-skills
/plugin install qa-testing@latestaiagents-agent-skills
# Skills-only (no MCP)
/plugin install safety@latestaiagents-agent-skills
/plugin install git-mastery@latestaiagents-agent-skills
/plugin install owasp-guardian@latestaiagents-agent-skills# Install full plugin directly
/plugin install latestaiagents/agent-skills/plugins/devops-sre
/plugin install latestaiagents/agent-skills/plugins/qa-testingAfter installing plugins, you get 20+ powerful slash commands:
| Command | What It Does |
|---|---|
/incident-response |
Start incident workflow with severity assessment |
/runbook |
Generate or execute operational runbooks |
/postmortem |
Create blameless postmortem documents |
/slo-check |
Check SLO/SLI compliance and error budgets |
/on-call |
On-call handoff and escalation workflows |
/k8s-debug |
Debug Kubernetes pods, deployments, services |
| Command | What It Does |
|---|---|
/build-rag |
Design and implement RAG pipelines |
/rag-debug |
Debug RAG retrieval quality issues |
/eval-rag |
Evaluate RAG system with RAGAS metrics |
| Command | What It Does |
|---|---|
/security-scan |
Run security analysis on code |
/fix-vulnerability |
Get remediation for vulnerabilities |
| Command | What It Does |
|---|---|
/git-undo |
Recover from any Git mistake |
/debug-this |
Structured debugging workflow |
| Command | What It Does |
|---|---|
/write-test-cases |
Generate comprehensive test cases |
/test-plan |
Create test strategy documents |
/bug-report |
Structure actionable bug reports |
| Command | What It Does |
|---|---|
/write-job-description |
Create inclusive job descriptions |
/performance-review |
Write effective performance reviews |
| Command | What It Does |
|---|---|
/design-agent |
Design multi-agent architectures |
/debug-agent |
Debug agent workflows and state |
Usage:
/incident-response
The command guides you step-by-step through the workflow.
Skills activate automatically based on what you're doing:
| You Say | Skill Activates | What Happens |
|---|---|---|
| "I have a merge conflict" | merge-conflict-surgeon |
Systematic conflict resolution workflow |
| "Run the database migration" | migration-safety |
Checks existing data, creates backup, asks confirmation |
| "Delete all the old log files" | file-operation-safety |
Shows what will be deleted, asks for explicit confirmation |
| "Help me design a multi-agent system" | agent-supervisor-pattern |
Provides architecture patterns and code templates |
| "Why is this API so expensive?" | token-cost-analyzer |
Analyzes token usage with current pricing |
| "Build a RAG system" | hybrid-retrieval |
Vector + keyword search with reranking |
| "My RAG gives wrong answers" | corrective-rag |
Self-healing RAG with document grading |
| "Implement GraphRAG" | graphrag-patterns |
Knowledge graph integration for multi-hop reasoning |
| "Is this code secure?" | secure-code-review |
Security-focused code review methodology |
| "Check for SQL injection" | injection-prevention |
SQL injection detection and prevention |
No special commands needed — just describe what you're doing.
Check for skill updates and keep your installation current:
# Check for available updates
npx skills check
# Update all installed skills to latest versions
npx skills update
# List your installed skills
npx skills listagent-skills/
├── skills/ # Cross-platform skills (via skills.sh)
│ ├── safety/ # Essential safety (5 skills)
│ │ └── ai-safety-guardrails/
│ ├── developer/ # For developers (19 skills)
│ │ ├── git-mastery/
│ │ ├── code-intelligence/
│ │ └── debug-detective/
│ ├── mlops/ # For architects (20 skills)
│ │ ├── agent-architect/ # Multi-agent + LangGraph (13 skills)
│ │ └── llmops-guardian/ # LLM operations (7 skills)
│ ├── rag-architect/ # RAG systems (7 skills)
│ └── security/ # Security skills (16 skills)
│ ├── owasp-guardian/ # OWASP Top 10 (10 skills)
│ └── common-security/ # Common practices (6 skills)
│
├── plugins/ # Claude Code / Cowork plugins (with MCP)
│ ├── devops-sre/ # DevOps/SRE (8 skills, 6 commands)
│ │ ├── .claude-plugin/
│ │ ├── .mcp.json # Datadog, PagerDuty, AWS, K8s
│ │ ├── commands/
│ │ └── skills/
│ ├── rag-architect/ # RAG (7 skills, 3 commands)
│ │ └── .mcp.json # Pinecone, Weaviate, Neo4j
│ ├── security-guardian/ # Security (10 skills, 2 commands)
│ │ └── .mcp.json # Snyk, Semgrep, Trivy, Vault
│ ├── agent-architect/ # Multi-agent (13 skills, 2 commands)
│ │ └── .mcp.json # LangSmith, Redis
│ ├── developer-toolkit/ # Developer (19 skills, 2 commands)
│ │ └── .mcp.json # GitHub, Jira, Sentry
│ ├── hr-people-ops/ # HR (6 skills, 2 commands)
│ │ └── .mcp.json # Greenhouse, Lever, Workday
│ └── qa-testing/ # QA (6 skills, 3 commands)
│ └── .mcp.json # TestRail, BrowserStack
│
├── commands/ # Base slash commands
├── .claude-plugin/ # Root plugin configuration
│ ├── plugin.json
│ └── marketplace.json # All plugins listed here
├── CLAUDE.md # AI context file
└── README.md
Works with any AI agent that supports skills.sh:
- Claude Code (Anthropic)
- Cursor
- Codex CLI (OpenAI)
- Windsurf
- Cline
- Aider
- Continue
- And 30+ more...
Found a bug? Have a skill idea? Open an issue.
See CLAUDE.md for development guidelines.
Plugins include optional MCP (Model Context Protocol) configurations. Before enabling:
- Review credentials: Only connect tools you trust
- Use environment variables: Never commit API keys to version control
- Test environments: Be careful with production credentials
- Audit configs: Review
.mcp.jsonfiles before copying
All MCP integrations are marked optional: true — only enable what you need.
- Skills run locally — no data sent to external servers
- MCP servers connect only to services you configure
- No telemetry or tracking in this repository
- Claude Code Plugins: code.claude.com/docs/en/plugins
- Skills.sh: skills.sh
- MCP Protocol: modelcontextprotocol.io
USE AT YOUR OWN RISK.
These skills are provided as educational resources and best-practice guidelines for AI agents. By using these skills, you acknowledge:
- No Warranty: Skills are provided "as is" without warranty of any kind.
- No Liability: Authors shall not be held liable for any damages or data loss.
- Your Responsibility: Test in safe environments, maintain backups, review AI actions before execution.
- AI Behavior: AI agents may interpret skills differently. Always confirm before destructive operations.
- MCP Integrations: Connecting to external services is at your own risk. Secure your API keys appropriately.
MIT License — See LICENSE file.
Use freely in personal and commercial projects.