Transcript Extraction

Turn transcripts into durable, evidence-linked extractions; this is a capture pass, not planning or implementation, and it preserves each claim's actual certainty.

Inputs

A. Transcript sources

A required selection of one or more user-supplied docs/transcripts/<day>/<order>-<slug>/ directories or their direct transcript.txt files, resolved from the repository root. <order> is a two-digit conversation sequence. Each directory must contain exactly one source named transcript.txt; sources are read-only.

B. Current repository record

Repository-supplied context documents, the initial idea, relevant ADRs, and verified product behavior that bear on transcript claims. Individual records are optional when absent; every existing relevant record is required context and authoritative for reconciliation.

C. Report template

The required bundled references/report-template.md, which defines every extraction report's required shape and empty-section notation.

Outputs