fruafr/skills

This repository holds a small collection of agent skills, which are folders of instructions and reference files that an AI coding tool like Claude Code can load automatically to get better at a specific specialized task. The idea is that instead of relying only on general knowledge, the AI reads the files in a skill folder and follows them when a matching task comes up. Right now the repository contains one skill called incus-doc, which gives offline, verified command-line help and documentation for Incus, a system for managing containers and virtual machines. The goal of this skill is to reduce mistakes and made-up answers by having the AI check its claims against the reference files included in the skill rather than guessing from memory, and it is kept up to date with Incus version 6.0. To use a skill from this repository, you copy the contents of its folder into the correct skills directory that your particular AI harness looks in, such as a project-level or global Claude Code skills folder, or an equivalent folder used by other compatible agent tools. Each skill folder is self contained, meaning it includes everything needed to run, such as a main instructions file plus any supporting reference documents or scripts. The repository also includes a template folder with a starting point for writing new skills, based on an open format for agent skills, extended with sections for things like decision rules, edge cases, and examples. Skills in this repository are released under the Apache 2.0 license by default, with each skill folder specifying its own license file. The author notes these skills are meant for demonstration and learning, and recommends testing them carefully before relying on them for anything important, since AI behavior can vary.