agentleFS
LogicOberLogicOber/Worldview-Skills/skills/film-video/cinematic-film/SKILL.md

cinematic-film

Skill2 starsChanged 53 days agoApache-2.0View on GitHub

Entry point for making a short film (2–5 minutes) from an idea. Routes through direction, story, script, locked assets, per-block prompts, parallel generation and review. Use when the user wants a short film, a narrative video, a product film with a story, or asks to turn a premise into a finished video.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every character, location and prop locked so they stay consistent across cuts.

    line 24
  2. Very specific:

    Read core-engine/film-pipeline/SKILL.md for the stage contract.

    line 66
  3. Distinctive:

    Say what was skipped: If TTS is not configured, if a model is unavailable, if a reference URL failed — it goes in the report, not in …

    line 77
  4. Very specific:

    No story, just a product: product-demo

    line 94

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Cinematic Film

Make a short film from one idea. This is the full pipeline; everything else in the repository is a specialised branch of it.

What the user gets

  • A finished video, 2–5 minutes, in the visual style they picked from reference cards.
  • Every character, location and prop locked so they stay consistent across cuts.
  • A script with a real dramatic spine — want, obstacle, loss, dilemma, change — because a film without one reads as a demo.
  • Dialogue and narration with stable cast voices, prepared recordings, and reviewed audio-driven lip sync where the provider supports it.
  • A folder they can rebuild from: direction, story, script, prompts, task IDs, review notes, locks.

What the user has to do

  1. Say what the film is about, in any number of words.
  2. Pick a visual style by letter when shown 3–6 reference images.
  3. Approve the outline (one screen: logline, conflict table, block map) and the script (one screen per block).
  4. Choose takes by number when generations come back.

That is the whole interface. Camera, lighting, pacing, palette, shot size, action density, voice timing and segment chaining are decided by the skill and stated, not asked.

How it runs

user premise
   │
   ▼
film-style-library   → style.json (one letter from the user)
   ▼
film-direction       → direction.md (intent, genre defaults, palette per location)
▼
film-story→ story.md (spine, conflict ladder, loss, dilemma, payoffs)   ← gate: user approves
   ▼
film-script  → script.md (blocks ≤30 s, every shot justified, build list) ← gate: user approves
   ▼
film-character-sheet ┐
film-location   ├ all assets for the whole film, in parallel, then locked
film-prop-sheet      │
film-screen-capture  ┘
   ▼
film-shot-prompt     → one prompt per block; calls film-action-combat for fights,
 film-dialogue-voiceover for lines
   ▼
film-generate-review → block 1 first (style check), then all blocks in one wave,
    review by beat, lock by index, assemble, continuity ledger
   ▼
audio pass    → music after picture lock

Read core-engine/film-pipeline/SKILL.md for the stage contract. Read each stage skill when you reach it. Do not work from memory.

For any spoken script, recommend ElevenLabs MCP before video generation; API, compatible third-party TTS, or supplied recordings also work. Run film-dialogue-voiceover to fix each character/narrator's real voice ID, produce and measure speech, then attach it with the visual references through a verified audio-capable interface. Resolve missing speech capability before a voiced batch. Music and final mixing can follow picture lock.

Rules this entry enforces

  • No video before the gates. Direction, story and script are approved before a single frame is generated. The only exception is the user saying "just generate it", in which case the skipped gates are written into review.md.
  • Whole-film assets before any block. The script's build list is generated as one parallel batch and locked. Generating assets scene by scene is how style drifts.
  • Block 1 as style check. One block, two to four takes, user confirms the look before the rest is fired.
  • Segments chain. Block N+1 opens on block N's last frame (first_frame) with identical reference_image URLs. Prompts carry the previous exit state in their first line.
  • Lock by index. Takes are presented as block-03 c2; the user answers with numbers; locks.json records why.
  • Say what was skipped. If TTS is not configured, if a model is unavailable, if a reference URL failed — it goes in the report, not in silence.

Files this skill produces

films/<slug>/
  style.json  direction.md  story.md  script.md  continuity.md  asset-map.json
  assets/{characters,locations,props,screens}/
  prompts/block-NN.txt
  takes/block-NN/{tasks.jsonl, review.md, locks.json, *.mp4}
  scenes/block-NN.mp4  scenes/block-NN-last.png
  final.mp4

Related entries

  • Shorter, single-scene work: game-cutscene-generator, anime-action-scene
  • No story, just a product: product-demo
  • Vertical: social-media-video

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.