// Agents and skills //

Data · runs in Claude Code

It takes the project's word-by-word transcript and, when they exist, the script and the research sheet — from which it builds the names database (`roteiro/nomes.json`) that you review by hand. It outputs `transcricao/mapa/` with context blocks, people mentioned (matched by sound, which catches names Whisper misspelled), objects, evidence, and places, numbers and dates with their timestamps, the words-per-minute clock, repetitions, stutters, filler words, and silences, and the speech × script comparison showing what was improvised and what stayed on the page — all of it also in `mapa.md`, to read in one go. The script measures, counts, and matches; the interpretation (who matters, what the evidence proves, what goes into the video) is Claude's, together with you. For people about to do the rough cut, build a media list, or fact-check, who first need to know what was said.

• Data
• Runs in Claude Code
• Intermediate