| name | split-clips |
| description | Split story text into ordered clip units after character, location, and prop analysis are ready. |
Split Clips
Purpose
Convert raw story text plus upstream analysis artifacts into an ordered set of clip boundaries that can be fed directly into screenplay generation.
Use This Skill When
- Character, location, and prop analysis artifacts are already available.
- The workflow needs deterministic clip units before screenplay generation.
- Story text changed and previous clip boundaries are now stale.
Do Not Use This Skill When
- Upstream analysis artifacts are missing.
- The caller wants screenplay output directly.
- The story is too incomplete to support stable clip segmentation.
Inputs
- Raw story text.
analysis.characters
analysis.locations
analysis.props
- Skill-owned prompt files selected by locale.
Outputs
- Ordered clip boundaries and clip metadata.
- Persistable clip split payload for the
clip.split artifact.
- UI summary of total clips and any warnings about boundary ambiguity.
Artifact Contract
- Reads:
story.raw, analysis.characters, analysis.locations, analysis.props
- Writes:
clip.split
- May invalidate downstream screenplay and storyboard artifacts.
Execution Rules
- Must run after character, location, and prop analysis.
- Preserve source order; clip indices must be stable and monotonic.
- Every generated clip must map back to real source text.
- Do not silently repair unresolved boundary mismatches.
Failure Conditions
- Any required upstream analysis artifact is missing.
- Model output is invalid JSON or fails schema validation.
- Clip boundaries cannot be matched back to the source text.
- The result contains zero clips or overlapping/unsorted clips.
Quality Bar
- Clip boundaries must be stable enough for reruns and downstream screenplay generation.
- Each clip should represent a coherent beat or scene unit.
- The skill must fail fast on invalid segmentation instead of returning fuzzy output.