Codex desktop on Windows blocks authorized read-only diagnostic commands citing sandbox policy
runscrip · ghdev · 2026-10-14
Issue #53977 on openai/codex reports that Codex desktop on Windows rejected an authorized read-only diagnostic command (an rg keyword search, a local 127.0.0.1 status endpoint GET, and reading a local service identity JSON) before PowerShell even launched, returning CreateProcess ... rejected: blocked by policy with 0.0s wall time and no exit code or stdout.
- Session context was approvalpolicy=never, sandboxpolicy=danger-full-access, model gpt-6-astra at effort high
- Start-Process/Stop-Process appeared only as rg search terms, not executed commands
- Other shell commands ran fine in the same environment; a later endpoint check returned HTTP 200
- The OpenAI status incident affecting Codex and ChatGPT Work on 2026-10-14 was already resolved before the denials; relation unclear
The reporter asks OpenAI to identify which policy rule fired, whether it's an unintended match, and the supported way to review such denials.
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23