Degardis

Degardis turns a structured AI skill source into a portable skill bundle. You write what your skill knows as small Markdown files and say which classes of work need which pieces. Degardis checks the structure and assembles one complete page per class of work, so the agent reads that page and starts.

Use Degardis when you want a skill to be easier to review, maintain, and rebuild than a hand-written collection of instructions.

Start here

You need Python 3.10 or newer.

python -m pip install degardis
degardis init my-skill
degardis validate my-skill
degardis build my-skill --output .artifacts

init writes a source that already compiles. build runs the same checks as validate and writes the bundle only when they pass. Getting started walks through the loop, including the repository's worked example.

What you write

Each is one Markdown file with YAML frontmatter. The reference manual has one topic per construct.

What you get

Each task becomes one page carrying everything it needs: its approach, the knowledge it names, and whatever that knowledge requires. Choices only the running agent can make stay with the agent: which tasks a request asks for, whether a hand-off applies, which facets the situation calls for, and whether a principle or guide applies.

All subject guidance is yours. Degardis adds only domain-neutral instructions for routing, reading, and conformance, and it never paraphrases, merges, summarizes, or drops what you wrote.

Commands

Run degardis COMMAND --help for the exact options. Only init and build write files.

Documentation

What a passing build does not tell you

Compilation establishes artifact integrity: the source compiles, the pages are complete, every link resolves. It does not establish that the skill works. Whether an agent holding the bundle does better work than one without it is a behavioral question, answered by using the skill against real situations — not by the compiler.

Degardis Authoring is a companion skill for agents that help create Degardis sources.