openai / openai/codex

Codex cannot reliably invoke or control Chrome extensions during extension debugging

Open
#44,162 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

browser bug model-behavior
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

What issue are you seeing?

I want to use Codex to debug a Chrome extension, but Codex repeatedly prompts me to perform browser/extension operations manually instead of invoking and controlling the Chrome extension itself.

In this workflow, Codex is unable to reliably interact with the Chrome extension UI or execute the browser-side actions needed for debugging. Even when the extension is installed and available in Chrome, Codex often falls back to instructions such as asking me to click buttons, open extension pages, or perform steps manually.

This makes end-to-end Chrome extension debugging impractical because the agent cannot autonomously inspect and operate the extension as part of the debugging loop.

Steps to reproduce
  1. Open a Chrome extension project in Codex.
  2. Ask Codex to debug or test the extension in Chrome.
  3. Have the extension installed/loaded in Chrome.
  4. Ask Codex to inspect the extension behavior, interact with its UI, or reproduce a bug.
  5. Codex repeatedly asks the user to perform the required Chrome/extension actions manually instead of invoking and controlling the extension itself.
Expected behavior

Codex should be able to invoke and control Chrome extensions normally when the Chrome integration/browser-control tooling is available, including:

  • opening and interacting with extension pages or popups;
  • triggering extension UI actions;
  • inspecting the resulting browser behavior;
  • performing the necessary steps itself during a debugging session instead of repeatedly handing them back to the user.

If a particular extension surface cannot be controlled because of a technical or security limitation, Codex should clearly identify the exact limitation rather than generically asking the user to perform the operation manually.

Actual behavior

Codex frequently falls back to manual-operation instructions, which breaks the autonomous debugging workflow and prevents reliable end-to-end testing of Chrome extensions.

Impact

This significantly limits Codex for Chrome-extension development. Extension debugging often requires repeated interactions between source changes, extension reloads, popup/options-page actions, page interactions, and inspection of the resulting state. Requiring the user to manually perform those actions defeats much of the value of agentic debugging.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No source file or test is named. Begin by locating the Chrome integration or browser-control tooling, then reproduce the workflow with an installed extension and record which extension surfaces require manual actions. Done means Codex can perform the required extension interactions or clearly reports a specific technical or security limitation.

Written by the indexing model from the issue text.

Assessment

Domain
devtools, web-dev
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.