What it does This repo ships 18 “skills” — curated, LLM-optimized knowledge bundles for core AWS services (IAM, Lambda, DynamoDB, EKS, Bedrock, etc.). Each skill is a local markdown file with patterns, CLI references, troubleshooting, and best practices. Install them into Claude Code or Codex CLI, and the agent reasons about AWS without hitting live APIs or streaming massive documentation.

The interesting bit The author draws a deliberate line against MCP servers: no extra infrastructure, no credential exposure, no latency from fetching AWS docs mid-conversation. The trade-off is static knowledge over live data, with a weekly auto-check against AWS documentation to keep content from going stale. It’s a bet that most AWS questions are reasoning problems, not API lookup problems.

Key highlights

• 18 services covered, from IAM to Bedrock, each with a standardized SKILL.md structure
• Explicitly designed for token efficiency and small context windows
• Weekly automated checks against AWS docs for content freshness
• Supports both Claude Code plugins and Codex CLI skill installer
• Includes metadata on last update so you know how current a skill is

Caveats

• The “weekly auto-check” claim is stated but not demonstrated in the repo; no CI config or update bot is visible in the README
• Individual skills must be installed separately in Codex CLI, which is more friction than the Claude Code marketplace flow
• Static knowledge means no live pricing, resource status, or API schema validation

Verdict Worth a look if you spend hours in Claude Code writing CloudFormation or debugging IAM policies and want the agent to stop hallucinating AWS edge cases. Skip it if you need real-time resource inspection or already have a working MCP setup you’re happy with.