Skip to content

Fix agent safety and reliability regressions - #255

Merged
esokullu merged 2 commits into
mainfrom
codex/fix-agent-safety-reliability
Aug 13, 2026
Merged

Fix agent safety and reliability regressions#255
esokullu merged 2 commits into
mainfrom
codex/fix-agent-safety-reliability

Conversation

@esokullu

Copy link
Copy Markdown
Owner

Summary

  • harden iframe targeting, exact-frame geometry claims, upload boundaries, CAPTCHA identity checks, and rich-text retry guards
  • make tool parsing, provider capability detection, completion handling, fetch fallback, loop detection, and planner localization fail safely
  • restore quota-aware tab-chat persistence, screenshot staging, and teacher-capture acknowledgement
  • keep Chrome and Firefox behavior mirrored and add focused regression coverage

Why

A review of the recent agent changes found several cases where malformed model output, ambiguous iframe state, transient provider failures, or storage pressure could cause unsafe dispatch, incorrect state, or lost UI/session data. The fixes address the root causes at validation and state-transition boundaries instead of masking downstream symptoms.

Impact

Agent actions now remain source- and origin-bound, malformed calls cannot dispatch with missing arguments, provider and persistence failures recover predictably, and browser-specific implementations retain parity.

Validation

  • npm test — 33 toolbar guard tests, 1,676 core tests, and 60 security checks
  • npm run test:fixtures — 165 browser fixture tests
  • npm run test:ci — 14 CI scenarios and cloud capture
  • git diff --check origin/main...HEAD

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 105a28c923

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread src/chrome/src/agent/rich-text-toolbar-probe.js Outdated

Copy link
Copy Markdown
Owner Author

@codex pls review

@chatgpt-codex-connector

Copy link
Copy Markdown

Codex Review: Didn't find any major issues. Keep it up!

Reviewed commit: d2058b6248

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

@esokullu
esokullu merged commit 03bc994 into main Aug 13, 2026
5 checks passed
@esokullu
esokullu deleted the codex/fix-agent-safety-reliability branch August 13, 2026 07:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants