The Skills Barn
326 production-ready Agent Skills for the tasks language models can't do reliably from training alone.
Ten catalogs
Every skill here is non-agent-native: it covers something a language model can't do reliably from its own training — current threat feeds, live warehouse state, licensed geospatial data, or deterministic validation against a published standard.
DevOps & Infrastructure
Threat modeling, CI/CD, cloud security, IaC, and monitoring skills for infrastructure teams.
Systems Engineering
Requirements, architecture, verification, and safety-critical engineering skills.
Project Management
Planning, decomposition, resourcing, and delivery skills for agentic project work.
Geospatial & Mapping
Routing, geocoding, isochrones, terrain, imagery, and spatial-join skills.
Web3 & Financial Ops
Smart-contract audit and DeFi liquidity skills with strict funds-safety controls.
Observability & SRE
Incident correlation and alert-fatigue reduction skills for on-call engineers.
Data Engineering
Pipeline, warehouse, and data-quality skills that run on live warehouse state.
Localization & i18n
Deterministic i18n validation against CLDR, ICU, and Unicode standards.
Cybersecurity
Malware sandboxing, forensics, and authorized-scope security skills.
Software Engineering
Migration, spec compilation, and architecture-audit skills for dev teams.
How it works
A Skill is a folder with a SKILL.md file — instructions an agent loads when a task matches. Skills are free. Where one needs live data or heavy compute, it calls a metered API endpoint that has a free tier.
For developers
Clone a skill into Claude Code:
cp -r skills/devops-infrastructure/ssl-certificate-lifecycle-monitoring ~/.claude/skills/
For agents
Every skill is indexed in manifest.json and llms.txt. Metered endpoints accept agent-native x402 payment — an agent pays per call in USDC, no human in the loop, after a free tier.
Common questions
What is an Agent Skill?
A SKILL.md file — plain-language instructions an agent loads when a task matches its description.
What does it cost?
The skills are free and open source. Skills that need live data call a metered endpoint with a free tier — you only pay for the data behind the skill.
Why not just let the model do it?
Because these tasks depend on things a model can't recall: today's CVE feed, your live warehouse state, licensed map data, or an exact answer from a standard the model only approximates.