사용 방법
- 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
- 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
- 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/machina-exm/film-studio-skills --skill film-breakdown시작 전 확인
- 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
- 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.
Film Breakdown
Turn a script, treatment, or one-paragraph idea into a scene list and production-ready shot cards.
Locate the studio
Work inside the current studio root. Use /docs/breakdown.md for the scene list and /prompts/ for shot-card files. If either path is missing, ask whether to create the needed folder or document. Ask one question at a time and wait for each answer.
Read the source
Accept any of these inputs:
- a full script
- a treatment
- a one-paragraph idea
Preserve supplied dialogue verbatim. Do not fill story gaps with invented facts. If the input is incomplete, interrogate the user scene by scene. Ask only the next question needed to define the current scene, then wait. Finish the current scene before moving to the next one.
Build the scene list
Write the scene table in /docs/breakdown.md. Give every scene a stable scene ID. For each scene, record its summary, location, time of day, characters, props, shot-card file, and status. Keep asset tags and state variants exactly as supplied. Mark unresolved information plainly instead of guessing.
Create one shot-card file per scene
Create /prompts/scNN-shot-cards.md for each scene, using its stable scene number. Represent every shot through exactly three lanes and 22 fields. Do not add a fourth lane or merge fields.
Lane 1: identity
Use these nine fields in this order:
- Scene / shot ID
- Location
- Time of day
- Characters with asset tags and state variants
- Props
- Description
- Verbatim dialogue
- Running time
- Complexity
Lane 2: direction
Use these six fields in this order:
- Shot goal
- Task as a verb
- Dramaturgy
- Blocking
- Acting
- Style device
Lane 3: camera / edit
Use these seven fields in this order:
- Size
- Movement
- Lens
- Angle
- Cut type
- Pace
- Transition
The three lanes contain 9 + 6 + 7 fields: 22 total. Use a separate heading for each shot and a table for each lane. Keep the same field order in every shot.
Enforce clip boundaries
Give each clip one action only. Target 10–15 seconds per clip. If a card contains multiple actions, split it into separate shots. If a single action cannot fit the target duration, ask the user how to divide or simplify it.
Separate in-frame text
Remove signs, screens, titles, and other in-frame text from generation instructions. Create /docs/text-tasks.md if any such text exists. Record each text task with its scene / shot ID, exact required text, placement, and context. Leave a pointer in the shot card to the matching text task. Do not ask generation to render the text.
Validate the breakdown
Check that every scene row points to one shot-card file. Check that every shot has exactly the three lanes and all 22 fields. Check that every clip has one action and runs 10–15 seconds. Check that dialogue remains verbatim and all in-frame text is separated. List unresolved fields and continue asking about them one at a time.
Close the output
Summarize the files written and the scenes and shots created. End with this note: the prompt is written later from the card almost mechanically because the three column groups map one-to-one to prompt blocks.
