chat stopped as a precaution

What version of Codex CLI is running?

codex --0.153.4

What subscription do you have?

ChatGPT Pro 20x

Which model were you using?

No response

What platform is your computer?

No response

What terminal emulator and version are you using (if applicable)?

No response

Codex doctor report

What issue are you seeing?

Title: Repeated GPT-6 Astra precautionary stops in Codex CLI 0.153.4 on WSL

Please investigate repeated misalignment_policy_violation terminations in a scientific computing project. The CLI displays “Chat stopped as a precaution” and offers no detailed findings. Local terminal metadata records “Potentially unintended activity.”

Environment: WSL/Linux x86_64, standalone Codex CLI 0.153.4, OpenAI provider. The latest affected turn used gpt-6-astra with high reasoning. At subsequent diagnosis, current configuration was xhigh; this was not changed by the diagnostic procedure.

The user reports that these precautionary stops only began after switching to GPT-6. This is a reported timeline, not a controlled model comparison; no model-switch test was performed.

Affected project thread IDs and UTC timestamps:

• 01a07a76-53ca-7120-952e-49e96ff63c1c — 2026-09-07T06:26:47.364Z
• 01a07aa1-d70b-71c2-8ea5-427d7ad2f928 — 2026-09-07T07:35:38.134Z
• 01a07ad2-017c-7bd0-8d31-0784b5485834 — 2026-09-08T03:21:24.461Z

Latest affected turn: 01a07f00-d223-7153-a506-9580e86bd4e4.

The latest stop followed an archived-result diagnostic and report update. The last two shell commands exited 0. A limited read-only review found the pasted aggregation change consistent with the existing evaluator; the report retains the failed scientific outcome. This is not a full audit of the entire session.

Read-only strict config parsing and native skill discovery succeeded: 31 local user skills, 6 system skills, no duplicate or shared-source runtime entries. No custom provider definitions or model instruction override files were found. Running CLI binaries resolve to the same version directory. No stopped task was replayed, and no safety setting was changed during diagnosis.

Can you identify the triggering behavior and determine whether these events were false positives or a client/service issue? Full transcripts, credentials and scientific datasets are intentionally not attached. Please advise what additional minimal evidence is required.

Status: DRAFT ONLY; not submitted.

What steps can reproduce the bug?

Uploaded thread: 01a07cd1-f1f2-7ee0-8e1b-ff11bcc9e069

What is the expected behavior?

No response

Additional information

No response

Start by reviewing the uploaded thread 01a07cd1-f1f2-7ee0-8e1b-ff11bcc9e069 and the listed affected thread IDs, including their UTC timestamps and CLI 0.153.4 context. Done means identifying the triggering behavior or determining that the evidence is insufficient to distinguish a false positive from a client or service issue, and documenting the minimal additional evidence required.