Files
codex-oauth-plus-onboarding…/skill/codex-oauth-plus-onboarding/references/mail2925-explicit-refresh-and-visible-confirmation-2026-05-11.md
T

1.9 KiB

2925 explicit refresh and visual confirmation lessons — 2026-05-11

Context

During a live OpenAI/Codex signup test using a 2925 account-pool alias, BOSS challenged two assumptions:

  1. Whether the visible screenshot really showed the 2925 mailbox page.
  2. Whether checking the mailbox via route reload / DOM polling counted as refreshing for latest mail.

Confirmed behavior

  • CDP DOM reads showed the active 2925 account was chickliu2026@2925.com, but an initial macOS screencapture captured the wrong visible app/window because the actual 2925 Chrome target was not activated first.
  • Correct visual confirmation requires:
    1. Find the CDP target whose URL contains 2925.com.
    2. Call /json/activate/<target_id> or otherwise bring the Chrome target to front.
    3. tell application "Google Chrome" to activate.
    4. Capture both:
      • Page.captureScreenshot from CDP, and
      • macOS screencapture after activation.
  • For mail arrival, BOSS does not consider Page.navigate route reload alone enough. Explicitly click the visible 2925 刷新 button and log that it was clicked.

Live result

For current alias chickliu2026rigm1r@2925.com:

  • Inbox: explicit refresh clicked 3 times; no current alias, no OpenAI mail, no code.
  • Junk: explicit refresh clicked 3 times; had OpenAI/ChatGPT stale mail/code, but not the current alias.
  • Therefore the correct stop condition is: do not buy Luban SMS number until email verification succeeds.

Future runner requirements

  • Add a refresh_clicked=true field to 2925 polling JSONL.
  • Poll both Inbox and Junk, but only accept candidate codes from mail matching the current alias/local-part or a strict current-run timestamp/window.
  • For user-visible verification in Feishu, send the CDP screenshot rather than relying only on macOS screenshot; if sending macOS screenshot, activate the correct Chrome target first.