Skill Maker — skill-as-workflow scaffolder

Creates a new skill where deterministic logic lives in scripts/run.py and the model only interprets output.

Workflow

1. Get name (kebab-case), one-line description, trigger keywords.
2. Run the engine:

python scripts/scaffold.py <name> --description "..." --triggers "kw1,kw2" --base-dir .opencode/skills

3. Implement real logic inside the generated scripts/run.py (run(payload) function).
4. Put interpretation rules in the generated references/notes.md.
5. Restart opencode (config is not hot-reloaded), then test: python <name>/scripts/run.py '{"input":"example"}'.

Do not

• Do not hand-write SKILL.md from scratch — always use scaffold.py.
• Do not let the model compute what the script should compute.