Degardis is a command-line compiler for authors of portable Agent Skills. It turns structured YAML sources into self-contained bundles for Claude, Codex, Copilot, Cursor, Roo, and ChatGPT. Before writing an artifact, it validates the source, selected profiles, generated paths, and generated links.
Authoring stays separate from generated output: edit the source files, validate them, and rebuild either an installable folder or a distributable ZIP.
Degardis requires Python 3.10 or later. Install it from PyPI:
python -m pip install degardis
degardis --versionFor an isolated command-line installation, use pipx install degardis.
To work on Degardis from this repository instead:
python -m pip install -e .
degardis validate examples/structured-summary
degardis list examples/structured-summary
degardis build examples/structured-summary --output .artifactsThe successful build ends with Summary: 1 skill built as folder. and reports
the artifact path. The generated folder is ready to inspect or install:
.artifacts/structured-summary/
SKILL.md
agents/openai.yaml
references/
scripts/
assets/
Degardis writes only beneath the requested output root. Rebuilding a skill replaces that skill's existing folder and ZIP there, while preserving unrelated entries. See Artifact format before building directly into an agent's skill directory.
Add the example's optional profile or produce a ZIP:
degardis build examples/structured-summary --profile detailed --output .artifacts
degardis build examples/structured-summary --zip --output .artifactsexamples/structured-summary is the canonical public example used throughout
the guides and CLI help. It summarizes supplied material from any subject;
synthetic skills under tests/fixtures are compiler test data.
The Degardis skills catalog contains reusable skills authored with Degardis.
For the smoothest authoring experience, use Degardis Authoring. It guides a skill from initial design through review, validation, packaging, and installation while applying Degardis conventions across manifests, workflows, entries, profiles, scripts, and assets.
degardis list PATH [PATH ...]
degardis validate PATH [PATH ...]
degardis build PATH [PATH ...] --output PATH [--profile [SKILL:]PROFILE] [--zip]Run degardis COMMAND --help for exact options and examples. All commands
accept individual skill directories, directories containing skills at any
depth, or a mixture of both. This lets a source tree group skills into
subdirectories while building all of them with one command.
An unqualified profile name applies to every selected skill that defines it.
SKILL:PROFILE selects one owner, and all selects all available profiles.
Supplying explicit selectors replaces manifest defaults for that build.
For exact command behavior, profile selectors, source schemas, and the Python API, see the reference.
| Reader goal | Document |
|---|---|
| Build and install the example | Getting started |
| Understand the source and artifact models | Concepts |
| Create or modify a skill | Authoring guide |
| Look up commands, schemas, or the Python API | Reference |
| Inspect and install generated output | Artifact format |
python -m unittest discover -s tests -v
python -m degardis validate examples/structured-summary
python -m degardis build examples/structured-summary --profile all --output .artifacts