BuilderIO/skills

What it solves

This project provides a collection of composable "skills" designed to enhance the capabilities of AI coding agents (such as Claude Code, Cursor, and GitHub Copilot). It addresses common friction points in agentic workflows, such as the lack of visual planning, inefficient use of expensive LLM tokens, ambiguity in task completion, and the tendency for agents to guess instead of reading current documentation.

How it works

These skills are implemented as modular instructions and tools that can be integrated into an agent's environment. They can be installed via a CLI tool or as a Claude Code plugin. The skills range from simple status conventions (like /quick-recap) to complex integrations with external tools (like /visual-plan for MDX-based diagrams or /rewind for accessing local screen memory via the Clips Desktop app).

Who it’s for

Developers using AI coding agents who want to increase the reliability, transparency, and efficiency of their agentic development process.

Highlights

• Visual Planning and Review: Converts text plans and PR diffs into interactive visual recaps and diagrams using MDX.
• Resource Optimization: Orchestrates high-cost frontier models to handle judgment and planning while delegating heavy lifting to cheaper agents.
• Context Recovery: Integrates with Clips Desktop to allow agents to recover recent screen context and transcripts.
• Agent Governance: Includes tools for auditing other agents' work (/agent-watchdog) and resolving competing strategies (/plan-arbiter).
• Documentation First: Forces agents to perform web searches for authoritative documentation to prevent version drift and hallucinations.