EdgeSpeak Name Speakers

Turn an existing diarized transcript from anonymous labels into a traceable named copy. This is an identity-enrichment step, not speaker diarization: edgespeak-transcribe --diarize decides which voice spoke each segment, while this skill determines whether there is enough evidence to associate a voice cluster with a person.

Do not imply that EdgeSpeak acoustically recognizes identities. A source title or participant list supplies candidates; it does not prove which anonymous voice is which.

Inputs to confirm

• The diarized EdgeSpeak JSON. If only media is available, first use edgespeak-transcribe --diarize -o transcript.json.
• The best available participant metadata: a user-confirmed roster, official episode/show notes, or the original video/YouTube URL. Ask for the source URL when names matter and the local media does not carry enough metadata.
• A new output path, normally <stem>.named.json. Never overwrite the anonymous master transcript.

If the agent cannot access the source page, ask the user to paste its title, description, and participant roster. Do not download the media merely to obtain metadata; for a YouTube source, use the metadata-only flow in edgespeak-yt-download when needed.

Workflow

1. Validate the master transcript

The input must have a segments array and every segment must contain speaker with either an anonymous string such as speaker_0 or null. If no speaker keys exist, stop and request a diarized JSON; naming a plain transcript would be guesswork.