agentleFS
scasellascasella/codex-gpt-pro/gpt-pro/SKILL.md

gpt-pro

Skill4 starsChanged 7 months agoMITView on GitHub

Operate ChatGPT Pro through Browser Use/Playwright or Computer Use. Use when the user asks Codex to ask GPT Pro, prompt GPT Pro, use ChatGPT Pro, send a prompt to ChatGPT Pro/GPT Pro, or wait for a GPT Pro result. Checks whether Browser Use/Playwright and/or Computer Use are available, asks for a first-run default route when both are available, selects Pro, submits prompts, waits through long reasoning or streaming, and returns the completed answer while pausing for login, MFA, CAPTCHA, account-security prompts, and never exporting or saving cookies, session tokens, or auth credentials.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. thatmike1/claude-skills★ 10same name
    2. 199-biotechnologies/claude-skill-gpt-pro★ 3same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

The file

Show its rules and text

We split this skill into the rules it gives an agent

22 rules. About 0 are common, 16 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈16

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the in-app browser, or the user's already-authenticated default browser via Computer Use when needed, to send the user's prompt to …

    line 8
  2. Distinctive:

    Check which browser-control routes are available in the current Codex session before navigating:

    line 12
  3. Very specific:

    If tool availability is unclear and tool_search exists, search for Playwright/browser navigation tools and Computer Use desktop-control …

    line 15
  4. Distinctive:

    If neither route is available, report that this skill requires Browser Use/Playwright or Computer Use and stop.

    line 16
  5. Distinctive:

    If exactly one route is available, use that route.

    line 17
  6. Very specific:

    If both routes are available, check for a saved default route in preferences.json next to this SKILL.md.

    line 18
  7. Very specific:

    If the current user prompt explicitly requests playwright, Browser Use, in-app browser, Computer Use, or the default browser, use that as …

    line 19
  8. Very specific:

    After the user chooses, save the preference as JSON in preferences.json next to this SKILL.md, for example:

    line 23
  9. Distinctive:

    If the saved preference is missing, malformed, or names an unavailable route, ignore it for the current run and ask again when both routes …

    line 31
  10. Distinctive:

    If the preference file cannot be written, continue with the selected route for the current request and tell the user that the default was …

    line 32
  11. Very specific:

    Use the selected route from Route Selection.

    line 36
  12. Distinctive:

    If the selected route fails before prompt submission, try the other available route when it exists and explain the switch.

    line 37
  13. Distinctive:

    Use only the browser's existing authenticated session.

    line 38
  14. Distinctive:

    Never request, read, export, save, persist, or recreate passwords, one-time codes, recovery codes, cookies, localStorage, sessionStorage, …

    line 39
  15. Distinctive:

    Start a new chat unless the user explicitly asks to continue the current conversation.

    line 40
  16. Distinctive:

    Verify that Pro is selected, such as by checking for the Pro composer pill or the Pro model-picker state.

    line 42
  17. Very specific:

    Wait until the response is complete.

    line 45
  18. Distinctive:

    Treat Computer Use as a fallback for the user's real browser session, not as a way to harvest or persist auth.

    line 50
  19. Distinctive:

    Prefer a new tab/window for a new prompt unless the user asks to continue a visible conversation.

    line 51
  20. Distinctive:

    Use the accessibility tree to identify the composer, Pro pill/model selector, send button, stop-streaming control, and final assistant text.

    line 52
  21. Distinctive:

    Treat the answer as complete only after streaming controls disappear and the latest assistant text has stabilized.

    line 57
  22. Distinctive:

    If the browser loses authentication or asks for a private verification step, stop at that boundary and ask the user to take over in the …

    line 59

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

GPT Pro

Use the in-app browser, or the user's already-authenticated default browser via Computer Use when needed, to send the user's prompt to ChatGPT Pro and wait for the final answer.

Route Selection

  1. Check which browser-control routes are available in the current Codex session before navigating:
    • Browser Use / Playwright route: available when mcp__playwright__ browser tools or equivalent in-app browser tools are present.
    • Computer Use route: available when mcp__computer_use__ tools are present.
    • If tool availability is unclear and tool_search exists, search for Playwright/browser navigation tools and Computer Use desktop-control tools.
  2. If neither route is available, report that this skill requires Browser Use/Playwright or Computer Use and stop.
  3. If exactly one route is available, use that route.
  4. If both routes are available, check for a saved default route in preferences.json next to this SKILL.md.
  5. If the current user prompt explicitly requests playwright, Browser Use, in-app browser, Computer Use, or the default browser, use that as the current-run route even when a saved default exists.
  6. If both routes are available and no saved default exists, ask the user which route to use by default going forward. Treat this as a separate preference question, not as a request to reroute the current run. Ask it even when the current prompt explicitly requested a one-off route, and keep the one-off route for the current run unless the user says otherwise:
    • playwright: use Browser Use / in-app browser by default.
    • computer-use: use Computer Use with the user's already-authenticated default browser by default.
  7. After the user chooses, save the preference as JSON in preferences.json next to this SKILL.md, for example:
{
  "default_route": "playwright"
}
  1. If the saved preference is missing, malformed, or names an unavailable route, ignore it for the current run and ask again when both routes are available.
  2. If the preference file cannot be written, continue with the selected route for the current request and tell the user that the default was not saved.

Workflow

  1. Use the selected route from Route Selection. For playwright, navigate the Browser Use / in-app browser to https://chatgpt.com/ if ChatGPT is not already open. For computer-use, open https://chatgpt.com/ in the user's default browser and operate the visible UI.
  2. If the selected route fails before prompt submission, try the other available route when it exists and explain the switch.
  3. Use only the browser's existing authenticated session. If login, MFA, CAPTCHA, account verification, billing, rate-limit, or security prompts appear, pause and ask the user to complete them in the browser.
  4. Never request, read, export, save, persist, or recreate passwords, one-time codes, recovery codes, cookies, localStorage, sessionStorage, authorization headers, browser profile files, or other ChatGPT/OpenAI auth material.
  5. Start a new chat unless the user explicitly asks to continue the current conversation.
  6. Open the model selector and choose Pro / Research-grade intelligence.
  7. Verify that Pro is selected, such as by checking for the Pro composer pill or the Pro model-picker state. If Pro is unavailable, report that directly and do not silently fall back to another model.
  8. Enter the user's prompt exactly, preserving constraints and requested wording. Attach files only when the user supplied or requested them.
  9. Submit the prompt.
  10. Wait until the response is complete. Poll while the UI shows Stop streaming, Reasoning..., a progress indicator, or changing assistant text. GPT Pro can take several minutes; keep waiting unless the user interrupts or the browser reaches a clear error state.
  11. Return the latest assistant answer and the conversation URL. Include any visible error, timeout, or uncertainty about model selection instead of overclaiming.

Computer Use Fallback

  • Treat Computer Use as a fallback for the user's real browser session, not as a way to harvest or persist auth.
  • Prefer a new tab/window for a new prompt unless the user asks to continue a visible conversation.
  • Use the accessibility tree to identify the composer, Pro pill/model selector, send button, stop-streaming control, and final assistant text.
  • Expect this route to be more fragile than Browser Use because it depends on window focus, layout, zoom, accessibility permissions, and visible UI state.

Completion Checks

  • Treat the answer as complete only after streaming controls disappear and the latest assistant text has stabilized.
  • Extract the final assistant message, not hidden reasoning or details panels.
  • If the browser loses authentication or asks for a private verification step, stop at that boundary and ask the user to take over in the browser.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.