Scored across 24 tools
Most tools map to a distinct stage of the demo pipeline (probe vs record vs render vs compose vs qa) and the descriptions make the boundaries explicit, including the 'render = full pipeline' superset relationship. A few pairs are genuinely close — validate_storyboard vs lint_storyboard (schema contract vs pacing/preflight lint), frames vs stills (evenly spaced from video vs named markers from clean take), and probe vs record — but each description disambiguates them clearly.
Two conventions coexist: verb_noun for authoring/utility tools (get_storyboard_schema, validate_storyboard, init_demo, import_trace) and bare noun for pipeline stages (probe, record, voice, compose, gif, stills, frames, walkthrough), plus noun_verb ordering in job_status/job_list/job_cancel. It is readable and the bare nouns deliberately mirror CLI subcommands, but the mixed verb style and the job_* ordering break any single predictable pattern.
24 tools is on the heavy side of the rubric even for a full video pipeline, since authoring, validation, job control, post-render QA, review artifacts, and publishing are all folded into one surface. Each tool is substantive and non-trivial, but the count sits in the borderline band rather than a tightly scoped set.
Coverage is broad and lifecycle-complete: scaffold (init_demo/import_trace), author (guide/schema), inspect, validate/lint/doctor, run the pipeline, QA the MP4, export review artifacts (gif/stills/frames/walkthrough/embed), and file feedback, with job_status/job_list/job_cancel for async control. Minor gaps remain: no cleanup/delete or reset of a demo directory, and no publish step (deferred to external CI workflow).