promptdojo_

The AI-native design pipeline — concept to shipped MP4 in one workflow — step 8 of 8

ch 29 · programmatic design8/8
Checkpoint

One last thing before we move on. Same surface as a write step — but the lesson doesn't complete until this passes.

Final drill. You're consulting for five teams. Audit each one's pipeline against the canonical seven steps.

Write audit_pipelines(teams) that takes a list of team dicts and returns a dict mapping each team's name to a verdict dict shaped like:

{"score": int, "missing": list[str]}

Each team dict has a name and a steps list (same shape as the previous drill — each step has step and tool fields).

  • Use the same scoring rule: 15 points per UNIQUE canonical step covered (step name in canonical list AND non-empty tool), capped at 100.
  • missing is a list of canonical step names that were NOT covered. Preserve canonical order: ["concept", "reference-gen", "composition", "audio", "captions", "render", "review"].
  • Five teams run. Expected output:
acme:        score=100 missing=[]
bravo:       score=90  missing=['captions']
charlie:     score=75  missing=['audio', 'captions']
delta:       score=60  missing=['reference-gen', 'audio', 'captions']
echo:        score=45  missing=['reference-gen', 'audio', 'captions', 'review']

full-screen editor opens — close anytime to keep reading.

read, then continue.