| name | create-skill |
| description | Use when asked to create or revise a LazyAI skill source file while keeping generated tool-native skill surfaces derived, not hand-edited. |
Create Skill
When to Use
Use this skill when the user asks to:
- create a new reusable skill
- rewrite an existing skill's trigger or workflow
- add a skill to a managed source library
- turn repeated guidance into one canonical skill artifact
Do not use when the user is only brainstorming. Use skill-authoring when modifying an existing skill in place.
Rule
- A shipped LazyAI skill has one source of truth. In this repository that source is
packages/cli/library/skills/<name>.md. Generated .opencode/skills/<name>/SKILL.md, .claude/skills/<name>/SKILL.md, .pi/skills/<name>/SKILL.md, and Copilot .github/skills/<name>/SKILL.md (global ~/.copilot/skills/<name>/SKILL.md) files are outputs, not authoring locations.
Workflow
- Confirm WHAT, HOW, DON'T WANT, and VALIDATE.
- Choose a kebab-case name.
- Write or update the canonical skill source markdown with
name and description frontmatter.
- Keep the source concise and operational: trigger, workflow, constraints, and verification.
- Prefer one markdown file unless the source library already has an approved multi-file pattern.
- Recompile affected tool surfaces and inspect generated outputs.
- Run
lazyai-cli compile plus lazyai-cli doctor in a consuming project, or go test ./packages/cli/... plus go build ./packages/cli/... when changing LazyAI's embedded skill library or adapters.
Constraints
- One skill, one clear trigger.
- Do not hand-edit generated tool-native skill files.
- Do not reintroduce retired workflow/task/orchestrator runtime assumptions.
- Keep optional detail proportional; do not hide the main workflow in deep references.
Verification Checklist