meeting-note-ingestor-granolalisted

# meeting-note-ingestor-granola Thin connector around `meeting-note-ingestor` for Granola.ai users. Granola JSON has a recognizable shape (top-level `title`, `attendees`, `transcript` fields), but the portable ingestor expects either a transcript file path with auto-detection or an explicit `--format granola` flag. This skill smooths the entry point: pass `--granola-json <path>`, get back the same structured note. It's a routing layer, not a separate generator — same output shape as the portable ingestor. The MCP fetch step itself is documented as an orchestration pattern (Claude session, cron, or downstream consumer skill calls `mcp__claude_ai_Granola__get_meeting_transcript` and writes the JSON to a tmp file), not embedded here. Embedding MCP calls in a shell skill would couple the foundation repo to a specific MCP version and to the adopter's MCP availability. Adopters who don't use Granola never invoke this skill; they call the portable ingestor directly with their format. ## How adopters use this There are two paths depending on whether the Granola transcript is already on disk: ### A. Granola transcript already on disk ```sh ./from-granola.sh --granola-json /tmp/granola-meeting-abc123.json --output /tmp/note.md # → writes structured note at /tmp/note.md # (delegates to ../meeting-note-ingestor/ingest.sh with --format granola) ``` ### B. Granola transcript fetched via MCP at runtime The connector pattern: Claude (or an automation harness) fetches the transcript