wordpress-workspace-transcription-smoke-test

WordPress Workspace Transcription Smoke Test

Use this skill for direct endpoint testing of WordPress.com transcription.

Workflow

1. Use the existing script first.
   • Script: Tools/wpcom-transcribe.sh
   • Endpoint: POST /wpcom/v2/sites/{site}/ai/transcription
   • Required: bearer token or auth header, site ID/domain, and audio file.
2. Keep it dependency-free.
   • The script is POSIX shell plus curl.
   • Do not add jq, Python, local OAuth flows, or extra dependencies unless the user explicitly asks.
3. Choose the narrowest invocation.
   • Dictation: WPCOM_BEARER_TOKEN="$TOKEN" Tools/wpcom-transcribe.sh --site 123456 --file ./sample.mp3
   • Command mode: add --intent command --selected-text "Rewrite this."
   • Debug REST proxy: add --envelope.
   • Network debugging: add --verbose and optional --proxy <url>.
   • Custom context: add --app-context '<json>'.
4. Protect secrets.
   • Never print token values.
   • Prefer environment variables or redacted examples.
5. Interpret failures by layer.
   • Local file and MIME detection.
   • Auth header or token scope.
   • Site ID/domain and selected-site permissions.
   • Endpoint response body and HTTP status.
   • Proxy, timeout, or WordPress.com API availability.

Product Notes

• The app uses the same site-scoped endpoint for dictation and selected-text command mode.
• The selected site may provide a server-side wp_guideline skill with slug transcribe, so connected apps share the same spelling, cleanup, and style rules.
• The endpoint may create the Transcription guideline as a WordPress Guideline skill on first use.
• Saving artifacts is controlled by the save_artifact field in app requests and should create WordPress Guideline artifacts on the selected site.