stevehuang0115/crewly/drive-read

Drive Read

bash execute.sh --id 1AbC…                        # print content (text/CSV) as JSON
bash execute.sh --id 1AbC… --out /tmp/report.pdf  # save (binary decoded) and print metadata

A file over the skill output cap (60 KB) is too big to print: without --out you get a one-line {"truncated":true,...,"next":"...--out <path>..."} envelope. With --out the cap is skipped and the exact bytes are saved (a binary of any size); --out fails with reason: no_content / truncated_response rather than report success without a file.

Output

{"file":{"id":"1AbC…","name":"Q3 plan","mimeType":"application/vnd.google-apps.document","webViewLink":"https://…"},"contentType":"text/plain","encoding":"utf8","bytes":1234,"content":"…"}

For a Google Doc / Sheet / Slides prefer docs-read / sheets-read / slides-read — they keep structure (headings, cell grid, per-slide lines). Files over 10 MB are refused with a webViewLink to open instead.

Failures

{"success":false,"reason":"not_connected","hint":"<connect URL>"} (exit 1) when the owner has not connected Google Workspace. Run the google-connect skill to post the owner a one-tap card in Slack (never send them to Crewly settings or the Connections page — they are usually on a phone). Manual fallback: https://crewlyai.com/portal/integrations/google. A 403 with reason: "google_error" on a write means the file was not created by Crewly — the grant only edits files Crewly made (drive.file).

Choosing a Google account

Several Google accounts can be connected at once. Without --account the call uses the default one (the first you connected, or whichever the owner marked default). Name one explicitly when it matters:

bash execute.sh --account work@company.com ...

The account must be connected for this product — Google consent is per product (Gmail / Calendar / Drive), so an account connected only for Calendar cannot read Drive.