What problem does it solve? A model that just wrote a lecture deck is the worst judge of whether it is right, and confident invention — fabricated quotations, unverified claims, broken structure — can propagate into student-facing materials. This Skill runs a separate acceptance gate that re-derives claims from primary sources instead of re-reading the draft. ## Core Features & Use Cases - Mechanical checks: Runs check_deck.py to report timing-cue coverage against the 90-minute budget, speaker-notes presence, [src:] source coverage, bullet density, and title style. - Structural review: Walks the deck against the five-step ladder (Tension, Core concept, Mechanism, Realistic complexity, 2026 frontier) and flags any slide that presupposes a later step. - Claim verification: Opens each [src:] source to confirm it says what the slide says, recomputes numbers, and requires dated observations for living benchmarks. - Use Case: After building class 7 of the IME AI-Applied-to-Transportation course, ask the gate to review the deck; it returns PASS, PASS WITH FIXES, or RETURN with numbered per-slide directives, then updates tracker.md from built to gated. ## Quick Start Run the lecture gate on class7.qmd to check it against lecture_pattern.md and tell me whether it passes.
