Digest / February 19, 2026
Codex Issues Digest — February 19, 2026
38 new issues filed ·
21 resolved in openai/codex (UTC day).
New issues (38)
- Clarity on the "Simple Codex" harness in Terminal Bench 2.0?
- Clean Copy for Multiline and Soft-Wrapped Output
- request_user_input glitches out when checking /agents
- stream disconnected before completion: error sending request for url (https://chatgpt.com/backend-api/codex/responses)
- stream disconnected before completion: error sending request for url (https://chatgpt.com/backend-api/codex/responses)
- Feature request: add a PreCompact hook event
- False Positive: Cyber
- sonatype‑2026‑000159
- CVE‑2025‑15284
- Request‑body parsing weakness; could enable crafted payload bypass.
- Tooltip hover shows "cmd" instead of "ctrl" in linux
- Panic in AbsolutePathBuf::parent() under EMFILE (Too many open files)
- Configurable Max Questions in Plan Mode
- LLM output is not displayed properly when running codex cli in alacritty + tmux
- Allow custom model selection for subagents
- GPT-5.3-Codex is lazy and irrationally terrified of breaking anything
- Windows `ssh` 9.5.5.1 doesn't like CodexSandboxUsers on ~/.ssh/id_rsa
- Cant hot reload skills, skill change needs restart
- Missing required fields in Responses API input items when using non-Azure endpoints
- Codex confuses cwd with target worktree
- exec is not allowed to use "GPT-5.3-Codex-Spark" model returns an error
- app-server WS handshake fails with iOS URLSessionWebSocketTask (-1005) before initialize
- Withdrawn (duplicate #11819)
- Ability to set reasoning effort for automations
- Failed to apply patch consantly
- Communicate the human readable subagent name when spawning
- Codex CLI hard crashes on Alpine Container
- --sandbox workspace-write fails with read-only mount in Azure DevOps Pipeline CI despite correct flags
- Shortcut Conflict: Ctrl + J conflicts with Chinese IMEs (Simplified/Traditional toggle)
- make the arrow a underline highlight or diff color or hover effect when using codex resume --all
- Add "Sign in with Device Code" to handle `localhost:1455` OAuth callback issues
- Model thinking level resets to "medium (default)" when sending prompt in 0.104.0 with different thinking level configured
- Improved sandbox diagnostics
- Falling back from WebSockets to HTTPS transport
- Codex crash during cleanup session caused unlogged deletion of ~328K files
- unified exec command can disappear when an exploring exec cell is active which hides ran commands from end users and is very misleading
- Allow writable `.git`/`gitdir`
- Approval menu - next to permissions
Resolved issues (21)
- Flagged for high-risk cyber activity when not even using gpt-5.3-codex
- request_user_input glitches out when checking /agents
- Codex App: Set font for built-in terminal
- Feature request: add a PreCompact hook event
- sonatype‑2026‑000159
- Allow access of `/dev/urandom` from bubblewrap sandbox
- Clarity on the "Simple Codex" harness in Terminal Bench 2.0?
- Deprecation warning links to removed docs anchor docs/config.md#feature-flags
- Subagent configuration and orchestration
- Allow custom model selection for subagents
- Missleading warning message: "Your account was flagged for potentially high-risk cyber activity ..."
- Account being flagged for “potentially high-risk cyber activity”
- Withdrawn (duplicate #11819)
- Codex CLI: File suggestion bug (“no matches”) @meantianing
- GPT-5.3-Codex is lazy and irrationally terrified of breaking anything
- exec is not allowed to use "GPT-5.3-Codex-Spark" model returns an error
- Impossible to turn off chatgpt.com and openai.com connections
- Model thinking level resets to "medium (default)" when sending prompt in 0.104.0 with different thinking level configured
- Cant hot reload skills, skill change needs restart
- Codex crash during cleanup session caused unlogged deletion of ~328K files
- Falling back from WebSockets to HTTPS transport