read_breakdown

Read a Breakdown tab: sources (play sessions, hosted documents, file sets), excerpt moments, findings and open questions. view=overview lists sources, counts, finding summaries and questions. view=transcript returns one play session as compact timestamped lines (best first read). view=moments returns excerpt rows with captions, notes, tags and frames; findings/questions return complete items. Filter with sourceId, fromTime/toTime, tag, query, momentIds, findingIds, section. Page with nextCursor + expectedRevision. Details: /docs/mcp/read_breakdown.

A Breakdown tab records how a game (or any product) is studied. Sources are play sessions (screen recordings with speech transcribed into timed moments), documents (hosted rich-text documents: read them with read_board(sheetId, documentId) and edit with apply_document_operations) and files (one entry holding many uploaded assets; each file has a moment; fetch bytes with get_asset_downloads). Findings are the write-up; questions are gaps to cover next time. Start with view=overview (sources with kind and status, counts, finding summaries, questions). Then view=transcript with sourceId reads a play session as compact lines: [m:ss] speech, plus screen captions, notes, tags and frame IDs; page with nextCursor and expectedRevision (about 400 moments per page). view=moments returns structured rows; focus with sourceId, momentIds, fromTime/toTime, tag or query. view=findings returns full bodies (filter section/findingIds); view=questions lists open and answered questions. Frames are immutable image assets: fetch bytes with get_asset_downloads(ids=[frame.media.id]) or read the stable /api/boards/{boardId}/assets/{id} src. No media bytes are inlined.