feat(examples): add Manim explainer production

[Cordy27]

Summary

• add a complete examples/manim-explainer production example
• add three independent Manim scene sources and locked Python environment
• add the project-local @project/manim-showcase component
• integrate Seedance presenter video, semantic Script Moments, Manim media tracks and Film rendering
• document Manim authoring, external render handoff and media verification rules
• update example asset ignore rules to retain required source media while excluding generated outputs

Validation

• rendered and inspected the three Manim scenes
• ran the complete Seedance-to-Hypit production workflow
• verified the final portrait video output, audio track, frame rate and canvas dimensions
• ran git diff --check
• commit: 0a7a8801

Notes

Seedance generation requires the configured external Provider and accepted Build Results.
Credentials are not included in the repository.

[rponeawa]

Thank you so much for this contribution! A Manim workflow is a great addition, and treating Manim as an external pre-renderer whose output enters Hypit as ordinary video is the right direction. A few things need to change before we can merge.

Skill docs: too much repeated guidance

• manim.md restates guidance the skill already has. Probing is covered by hypit media probe in media.md, frame grids by review.md, and external file dependencies by project-files.md. Please link to those sections and keep only what is specific to Manim: the pre-render boundary, the locked uv environment, and the render command.
• The same Manim summary is now in four places: a SKILL.md paragraph, a SKILL.md table row, a new media.md section, and two additions to project-files.md. The table row plus manim.md is enough. Please remove the rest.
• The new SKILL.md paragraph was inserted in the middle of an existing paragraph and splits off the "Judge preparation by..." sentences.
• In manim.md, "The project-local lockfile fixes Python packages, native libraries or the FFmpeg executable" should say it does not pin native libraries or FFmpeg.

Example: cannot run from a fresh clone

• runs/render.svrun reuses bld_20260922T074050519Z_38D612AD42, which exists only on your machine. Please either provide a downloadable media/Result bundle like complex-explainer does, or drop the build-record entries so the Run generates from scratch.
• The three manim-renders/*.mp4 files are referenced but not committed, and no ignore rule covers them.
• The root render.svrun points to ./main.svml, which does not exist. It duplicates runs/render.svrun and can be removed.
• pnpm-lock.yaml has no entry for the new workspace package, so pnpm install --frozen-lockfile in CI should fail. Running pnpm install at the repo root and committing the lockfile should fix it.

Example: does not follow its own guidance

• The Manim videos go straight into the component without Normalize, which the new docs say is required.
• ASSET-PROVENANCE.md notes that two of the renders are not constant 30 fps, and render.ts hardcodes their frame counts (165, 143, 154). Normalizing the renders would fix both problems.
• The presenter picture comes from english-seedance-person-full.mp4, while the audio and Timeline come from the five aligned Seedance clips. Nothing keeps them in sync, so please use the normalized clips for the picture too.

Example docs

• The README, BRIEF, TREATMENT, CRAFT-NOTES and the component README all repeat the same "Manim owns internal animation, the component owns sampling, placement, scale..." paragraph. Once in the README is enough.
• ASSET-PROVENANCE.md mentions files that are not in the PR (english-seedance-native-voice.wav, drafts/legacy/, qa-*, final/).
• The pyproject.toml description still says "Taylor expansion explainer".
• The README render commands pass both -r and --resolution, which are the same flag.

Thanks again, and happy to help if anything here is unclear!