skills-library/AGENTS.md

AGENTS.md — Project context for AI assistants

Project objective

Build and maintain the Fedora AI Skills Library: a collection of reusable AI skills for Fedora community workflows.

Skills are self-contained, version-controlled directories that give AI agents specialized knowledge and workflows for specific tasks. The library is hosted publicly on Fedora Forge and is open to community contributions.

• Success looks like: A growing library of high-quality, reliable skills that Fedora contributors can load into any skills-compatible AI agent to automate or assist with community workflows.
• Non-goals: This library does not provide a hosted service, API, or bot. Skills are tools for humans to use with their own AI agents.

How we work (agentic first)

• Human role: Intent, skill design decisions, output quality review, final editing before merge.
• Agent role: Implementation drafts for skill instructions, reference files, templates, and documentation — always reviewed by a human.
• Contributions: Many Fedora contributors have write access. Contributions follow the workflow described in CONTRIBUTING.md.

Architecture

The library follows the Agent Skills open format. Each skill is a directory under skills/ containing a SKILL.md file with YAML frontmatter and instructions, plus optional supporting subdirectories.

skills/
└── <skill-name>/
    ├── SKILL.md          # Required: YAML frontmatter (name, description) + instructions
    ├── references/       # Optional: reference documentation loaded on demand
    └── assets/           # Optional: templates and static resources

Skill-specific context — workflows, conventions, reference data, and examples — lives inside the skill directory, not at the repo level. See each skill's SKILL.md for its own detailed instructions.

Repository layout

Path	Purpose
skills/	Skills directory — one subdirectory per skill
skills/<skill-name>/SKILL.md	Skill entry point: metadata + instructions
AGENTS.md	This file — project context for AI assistants
CONTRIBUTING.md	How to contribute
REDHAT.md	Red Hat AI policy reference
README.md	Primary project README
LICENSE.txt	CC BY-SA 4.0

Conventions

• License: CC BY-SA 4.0. All contributions are made under this license.
• Commits: Include Assisted-by: or Generated-by: per REDHAT.md when AI-assisted.
• Markdown: One sentence per line (ventilated prose). This produces cleaner diffs and easier review.
• Secrets: Never commit API keys, tokens, or credentials.
• Template placeholders: Use {{DOUBLE_BRACES}} syntax in template files.

Common tasks

• "Add a new skill — I want to add a skill for [workflow]. Create a new directory under skills/ with a SKILL.md following the Agent Skills format."
• "Improve an existing skill — I want to improve how the [skill-name] skill handles [specific aspect]. Read skills/<skill-name>/SKILL.md and suggest changes."
• "Update a skill's reference data — [something] has changed. Update the relevant file in skills/<skill-name>/references/."

Key links

• Red Hat AI policy: REDHAT.md
• Agent Skills specification: agentskills.io/specification