export-ml-project

Export ML Project

Ordering only. Children own convert and site commands.

Human-facing prose

Details: setup-workspace references/human_facing_prose.md. The multi-select uses Executed notebooks and Documentation site. Do not put skill ids, convert flags, or the wrapper CLI in the question.

Sequence

1. python -m skore_skills status. Read policy.notebooks, policy.site, and skills.

2. AskUserQuestion with allow_multiple. Say first, in 2–4 lines, what each box authorizes (toolchain installs, convert re-execution, a built site), the recorded flags (policy.notebooks and policy.site: true, false, or not recorded), and that both answers persist as policy. A file link is an addition, never the context. The question's last line is exactly: Select each one you want. Write nothing after that line.

   • Executed notebooks
   • Documentation site

3. After the answer, and not in the question, persist each box this turn: selected → python -m skore_skills policy set notebooks true (or site); a box the user does not select → false. Do not leave them null. Do not write notebooks or site into JOURNAL; policy is the record.

4. Load each selected child only if status.skills.<id> is true; else one-line skip. Do not invent that skill's steps:

   • notebooks → export-ml-notebook (installs toolchain if this is the first yes, then convert; --html only if the user wants the notebook viewer on the site)
   • site → export-ml-site (installs mkdocs-material if first yes, init if needed, then build; never convert)

5. Load triage-ml-task only if status.skills.triage-ml-task is true; else stop.

Stop conditions

• Do not git end-turn or git commit.
• Do not invent convert or mkdocs steps when the child skill is missing.
• Do not pixi add / uv add from this coordinator.