Skills Reference
Skills are specialized instruction files that the agent loads on demand when a task matches
their domain. They provide conventions, patterns, and guardrails for specific task types.
Each skill is a SKILL.md file that gets injected into the agent's context
when relevant.
Available Skills
| Skill | Domain | When Loaded |
|---|---|---|
| cleave | Task decomposition & assessment | Using /cleave, /assess, or cleave_assess |
| git | Git conventions | Committing, branching, tagging, releases |
| oci | Container & OCI artifacts | Containerfile authoring, registry work, multi-arch builds |
| openspec | Spec-driven development | Using OpenSpec commands (/opsx:*) |
| extensions | Extension API | Creating or modifying Omegon extensions |
| tui | TUI component patterns | Building extension UIs |
| python | Python development | Working with Python projects (pyproject.toml, pytest, ruff) |
| rust | Rust development | Working with Rust projects (Cargo, clippy, Zellij plugins) |
| security | Security checklist | User-facing code, template rendering, process spawning |
| style | Visual style guide | Creating visual output (diagrams, themes, images) |
| typescript | TypeScript conventions | Working with TypeScript (strict typing, node:test) |
| vault | Obsidian markdown | Generating interlinked documentation |
How Skills Work
- The agent detects that a task matches a skill's domain
- It reads the skill's
SKILL.mdfile using thereadtool - The skill's instructions are incorporated into the agent's reasoning
- The agent follows the skill's conventions for the duration of the task
Skills are loaded lazily — they only consume context when actually needed. The agent manages skill loading autonomously; operators don't need to specify which skills to use.