| name | analyze-locations |
| description | Extract normalized location analysis from story text before clip splitting. |
Analyze Locations
Purpose
Produce the canonical location analysis artifact that downstream clip splitting and screenplay generation rely on. This skill turns freeform setting mentions into stable structured location records.
Use This Skill When
- Raw story text exists and structured locations are missing.
- Story text changed and existing location analysis is stale.
- Clip splitting needs normalized place names before partitioning the story.
Do Not Use This Skill When
- The workflow only needs character or prop refreshes.
- Screenplay generation is the primary goal and location analysis is already current.
- The caller expects this skill to infer clip boundaries.
Inputs
- Raw story text.
- Existing base locations from project state.
- Skill-owned prompt files selected by locale.
Outputs
- Normalized location analysis rows.
- Persistable location analysis object for the
analysis.locations artifact.
- UI summary for newly identified, preserved, or merged locations.
Artifact Contract
- Reads:
story.raw
- Writes:
analysis.locations
- May invalidate downstream artifacts that depend on location analysis.
Execution Rules
- Preserve canonical project location names when a stable match already exists.
- Keep output strictly structured; do not emit screenplay prose.
- Merge equivalent location names when the evidence is clear.
- Reject malformed model output instead of patching it silently.
Failure Conditions
- Source text is empty.
- Model output is invalid JSON or fails schema validation.
- No usable location records are produced.
- The output introduces unresolved duplicate locations that cannot be normalized safely.
Quality Bar
- Each location must map back to real story evidence.
- The result should be concise, deduplicated, and stable for downstream reuse.
- Output must be deterministic enough that reruns do not churn names unnecessarily.
