Your AI agent's missing manual for prod infrastructure

What it does

This repo packages 160-plus operational skills—covering Kubernetes hardening, Terraform modules, SOC2 compliance, and LLMOps—into a standardized SKILL.md format that AI agents ingest. Each skill bundles real configs, CLI commands, troubleshooting steps, and cross-references so your agent stops hallucinating kubectl flags and starts citing actual production patterns. Humans can browse it too, but it is built for machine consumption first.

The interesting bit

The project rides the agentskills.io open format, treating tribal ops knowledge as versioned dependencies you install via npx skills add rather than static documents that rot in a wiki. It is essentially a package manager for expertise: your agent loads YAML frontmatter for matching, then pulls the full runbook only when you ask about, say, eBPF observability or MCP server security.

Key highlights

• 160+ skills across DevOps, security, cloud infrastructure, compliance, and AI engineering, each with 250–400+ lines of expert guidance
• Real, copy-pasteable configs in YAML, HCL, and Bash—not just links to external docs
• Covers emerging topics like AI coding agent guardrails, supply chain attack response, and LLM cost optimization
• Installs into Claude Code, Cursor, Codex, and other agents via the skills.sh CLI or a plain git clone
• MIT licensed, with scripts and templates meant for direct execution

Verdict

Grab this if you want your AI agent to stop winging infrastructure tasks and start following vetted SRE playbooks. Skip it if you are looking for a standalone automation framework—this is curated knowledge, not executable software.