gpt-pro-question-window

name	gpt-pro-question-window
description	Bridge Codex to a signed-in ChatGPT/GPT Pro conversation, including scoped file upload, conversation reuse, raw answer capture, and later Codex verification. Use for normal GPT Pro questions and as the browser/persistence foundation for other Codex Pro Bridge skills; do not use for local tasks that need no external reasoning.

GPT Pro Question Window

Use this skill as the browser and persistence adapter for Codex Pro Bridge.

Before creating or resuming bridge state, read references/bridge_protocol.md. It is the single source of truth for IDs, events, storage, invariants, and CLI commands.

Required flow

1. Identify the exact question, desired output, and one bridge-thread-id.
2. Reuse the GPT Pro session only when its local metadata points to the intended web conversation and the same bridge thread. Otherwise create a new task-scoped conversation.
3. Open ChatGPT in the user's signed-in Chrome session. Ask the user to handle login, passwords, 2FA, CAPTCHA, rate limits, or account-security prompts.
4. Upload a focused zip when evidence is needed. Never replace a failed upload with a full repository paste unless the user explicitly approves that fallback.
5. Read the exact selected model label and visible attachment name. Run scripts/check_browser_preflight.py before clicking Send. If the requested model is Pro, labels such as 极高 or an account name containing “Pro” do not satisfy the gate.
6. Wait until the response visibly finishes. Keep one submission active, report periodic progress during long waits, and never resend while generation is still active.
7. Immediately capture the prompt, bundle digest, full raw answer, model labels, attachment name, upload route, and observed timing with scripts/save_bridge_turn.py.
8. Re-open local evidence, verify the answer, and record the result separately with scripts/record_codex_verdict.py.
9. Run scripts/verify_bridge_thread.py --require-complete-rounds before a follow-up round or final handoff.
10. Report the saved turn and verdict paths, useful conclusions, rejected claims, and next action.

Completion criterion: the raw exchange and Codex verdict are separate immutable artifacts on the same thread, and every acted-on GPT Pro claim has a local verdict.

Normal-question prompt

For a normal question, read and use references/question_window_prompt.md. Specialized review skills provide their own prompt.

Chrome upload

Use Chrome's file chooser before Computer Use:

1. Before the first browser round, verify that the Codex Chrome extension is installed and enabled. In this environment, use a US-region network node while downloading it from the Chrome Web Store.
2. Open chrome://extensions/, open the extension's Details, and verify Allow access to file URLs is enabled.
3. Build the zip locally and keep its output path absolute for the browser call.
4. Start waitForEvent("filechooser") before clicking ChatGPT's visible attachment button and visible Upload from computer menu item.
5. Call chooser.setFiles([absolute_path]).
6. Verify the filename or attachment chip before submitting.
7. Remove the attachment and confirm an empty composer after a dry run.

Never directly click a hidden input such as #upload-files. Use a semantic visible control first; use programmatic file assignment only as a verified fallback.

Before submission, run:

python3 .agents/skills/gpt-pro-question-window/scripts/check_browser_preflight.py \
  --requested-model Pro \
  --selected-ui-label '<exact visible label>' \
  --bundle /absolute/path/to/bundle.zip \
  --attachment-name '<visible filename>' \
  --upload-control visible-menu

If the extension is missing or the permission is disabled, stop and fix that prerequisite before retrying. Use Computer Use only when Chrome still cannot control the native or graphical UI boundary after the extension and permission checks pass.

Treat attachment preprocessing that stalls before submission as a bundle-shape problem: regenerate a smaller package or at most two or three focused attachments. Do not interrupt a response that remains visibly active merely because a Pro run is slow.

Browser pacing

• Prefer one conversation. Use two or three only for independent tasks, never the same bridge thread concurrently.
• Wait for visible completion between upload, submit, streaming, and copy operations. During a long generation, inspect at roughly 30–60 second intervals and send concise progress updates at least once per minute.
• Distinguish submitted, generation observed, response complete, captured, and failed. Record observed timestamps; do not invent missing ones.
• If progress disappears, capture diagnostics and mark the attempt failed. Do not automatically resubmit the prompt.
• Inspect after failure; avoid rapid retries, scraping, or burst submission.
• Stop for service or account protections rather than attempting to bypass them.

Non-negotiable checks

• Keep uploaded evidence inside the user-approved scope.
• Save the answer before using it.
• Verify locally before editing code or trusting a result.
• Record model state as verified, mismatch, or unverified; never upgrade a mismatch to a Pro claim.
• Verify ledger parents, artifact hashes, and bundle hashes before continuing the thread.
• Never move an existing local session to another thread or web URL.
• Never overwrite a saved bundle, turn, snapshot, or verdict.