メニュー
Pull Google Drive files (recordings, docs) into the raw inbox via Drive MCP; metadata only — transcription is explicit via /dd:raw:transcribe. Folders come from raw.drive.* config. Standalone-invokable; also wrapped by /dd:raw:ingest-drive.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Capture an external signal (meeting note, email, idea, client request) into <repo>/<paths.raw_root>/_incoming/ with structured frontmatter. Universal — sources and filters come from <repo>/.drydock/config.yaml. Standalone-invokable; also wrapped by /dd:raw:capture.
Pull Google Calendar events into the raw inbox via the Calendar MCP. Calendars and time window come from raw.calendar.* config. Standalone-invokable; also wrapped by /dd:raw:ingest-calendar.
Pull issues and comments from external GitHub repos we collaborate on (not the current repo) into the raw inbox via gh CLI. Repos come from raw.github.repos. Standalone-invokable; also wrapped by /dd:raw:ingest-github.
Pull Gmail messages matching configured filters into the raw inbox via the Gmail MCP. Filters and look-back window come from raw.gmail.* config. Universal — no client-specific filters. Standalone-invokable; also wrapped by /dd:raw:ingest-gmail.
Pull Linear issues and comments from configured teams into the raw inbox via the Linear MCP. Teams and filters come from raw.linear.* config. Standalone-invokable; also wrapped by /dd:raw:ingest-linear.
Pull Notion pages from configured databases into the raw inbox via the Notion MCP. Databases come from raw.notion.databases — list of {id, label}. Standalone-invokable; also wrapped by /dd:raw:ingest-notion.
| name | raw-ingest-drive |
| description | Pull Google Drive files (recordings, docs) into the raw inbox via Drive MCP; metadata only — transcription is explicit via /dd:raw:transcribe. Folders come from raw.drive.* config. Standalone-invokable; also wrapped by /dd:raw:ingest-drive. |
You are the raw-ingest-drive skill. Bulk-ingest Drive file metadata as raw items. Recordings stay metadata-only — transcription is explicit and separate via raw-transcribe.
source "${CLAUDE_PLUGIN_ROOT}/lib/config.sh"; dd_config_load
source "${CLAUDE_PLUGIN_ROOT}/lib/hooks.sh"
source "${CLAUDE_PLUGIN_ROOT}/lib/frontmatter.sh"
MCP detection: Google Drive MCP must be connected. Absent → fail with install guidance.
Read config:
folders="$(cfg_array_get raw.drive.folder_ids)"
file_types="$(cfg_array_get raw.drive.file_types)"
since_days="$(cfg_get raw.drive.since_days)" # default 7
Empty folders → abort: "raw.drive.folder_ids must be configured."
Pull file metadata for files in each folder modified in the last since_days matching any of file_types. Do NOT fetch file bodies wholesale — metadata + a small preview only (first ~500 chars for Docs/Sheets).
For each file:
dedup_key: drive:<file-id>. Skip if exists.${modified_date:0:10}_drive_$(fm_slug "${file_name}").md.source: drive, captured_at, dedup_key, original_at: <modifiedTime>, parties: [owner], links: [webViewLink], attachments: [drive:<file-id>], drive.{file_id, mime_type, size_bytes, is_recording}, traces_to: {}.dd_hook_invoke post-capture with {item_path, source: "drive"}.Report: ingested / skipped, plus a count of recordings (so the user knows how many /dd:raw:transcribe invocations are pending).
raw.drive.* only.