AI Skills (aiskills)
Reusable, tool-agnostic AI engineering skills, workflows, and evaluation playbooks for coding agents
Overview
aiskills is an open-source collection of reusable, tool-agnostic AI engineering skills, workflows, templates, and evaluation playbooks for AI coding agents. Designed to be composable across Cursor, Claude, Windsurf, and other agent environments — covering RAG implementation patterns, LLM evaluation harnesses, agent workflow templates, and production AI engineering best practices.
Problem
AI engineers repeatedly rewrite the same RAG pipeline boilerplate, evaluation harnesses, and agent workflow scaffolding across projects. No tool-agnostic, composable library exists for sharing these as reusable skills.
Solution
A structured skill library where each skill is a self-contained, documented pattern: RAG chunking strategies, embedding pipeline templates, evaluation harnesses for LLM output quality, agent workflow blueprints, and prompt engineering playbooks — all tool-agnostic and composable.
Architecture
- 1Skill modules — self-contained patterns with documentation and usage examples
- 2RAG pipeline templates — chunking, embedding, retrieval, and citation patterns
- 3LLM evaluation harnesses — grounded output quality measurement playbooks
- 4Agent workflow blueprints — orchestrator + specialist patterns
- 5Prompt engineering playbooks — structured output, tool-calling, and persona patterns
- 6Tool-agnostic design — compatible with Cursor, Claude, Windsurf, and VS Code
Technical Challenges
- Designing skill abstractions generic enough to be tool-agnostic while specific enough to be immediately useful.
- Structuring evaluation harnesses that measure real LLM output quality without requiring live API calls in CI.
Results
- Published and actively maintained open-source AI engineering skills library.
- 1 fork, adopted by the AI engineering community for composable agent skill sharing.