Portable skills for AI agents that use 46 different formats

What it does

SkillKit is a CLI tool that acts as a package manager for AI coding agent “skills” — essentially prompt libraries and instruction sets. It detects which agents are present in your project, pulls skills from sources like GitHub or GitLab, runs a security scan, translates the files into each agent’s preferred format, and drops them into the correct directories. It also remembers session learnings, generates new skills from plain English, and exposes a local REST API so agents can discover skills at runtime.

The interesting bit

The auto-translation layer is the real workhorse: it takes a single skill and reformats it for each agent’s specific directory and file type, from SKILL.md for Claude Code to .mdc for Cursor. It even generates agent instruction files like CLAUDE.md and .cursorrules from your codebase, treating vendor-specific prompt formats as compile targets rather than source code.

Key highlights

• Supports 46 agents and 31 skill sources, including GitHub, GitLab, gists, and local paths.
• Auto-translates skills between agent formats and auto-generates agent instruction files.
• Stack-aware recommendations that read your repo and rank relevant skills.
• Optional add-on packages for a TUI, REST server, P2P mesh, and agent messaging; core stays slim at 118 packages.
• Session memory compression to capture what an agent learned so it doesn’t vanish when the context window closes.

Verdict

Worth a look if you juggle multiple AI coding agents or want to share prompt libraries across a team without standardizing on a single vendor. Skip it if you are happily locked into one agent and don’t mind maintaining your own .cursorrules by hand.