Forumics logo Forumics
Issue 01

Back to issue

All Codex known issues

CLAIM C054 Supported 2026-11-07

GitHub issue #50769 in openai/codex, opened 2026-10-04 by jabdori, records a report that during October 3-4, 2026, development coordinated through dots hit repeated pre-execution approval blocks although the user had authorized development and publication, that later denials kept citing read-only scope or missing authorization, that a branch-creation call returned a user-cancelled message the user said they had not triggered, and that moving the workflow to gh CLI on the connected Mac still left branch creation rejected before execution, on macOS with GitHub CLI 2.101.0 and unknown Codex, client and backend builds.

Verbatim Excerpts from Cited Records
“During October 3–4, 2026 (UTC), development coordinated through dots encountered repeated pre-execution approval blocks. The user had authorized development and publication, while separate scheduled progress reports were read-only. Later denials continued to cite read-only scope or missing authorization.”
— jabdori · openai/codex (issue 50769 opening body, opening paragraph) · 2026-10-04T02:36:36Z ·GitHub Issue ↗
“In a separate private-repository development workflow, tree creation and a PR-review request were denied with explanations citing the read-only reporting restriction and no later authorization. The user had directly requested development to resume after the reporting task.”
— jabdori · openai/codex (issue 50769 opening body, Incident timeline section) · 2026-10-04T02:36:36Z ·GitHub Issue ↗
“Around 01:55 UTC: The subsequent branch-creation call returned “user cancelled MCP tool call”. At 02:15 UTC, the user explicitly stated they had not cancelled anything. The actual cancellation source is unknown.”
— jabdori · openai/codex (issue 50769 opening body, Incident timeline section) · 2026-10-04T02:36:36Z ·GitHub Issue ↗
“At the user's request, the workflow moved to gh CLI on the connected Mac. Read-only authentication/repository checks succeeded. Branch creation was still rejected before command execution, citing the original read-only scope and treating forwarded later approval as insufficiently attributable. One retry with the original question and user response was also denied. No branch, draft PR or CI run resulted for this private-repository candidate.”
— jabdori · openai/codex (issue 50769 opening body, Incident timeline section) · 2026-10-04T02:36:36Z ·GitHub Issue ↗
Counter-record: Issue remains open with 32 comments as of 2026-11-05; no maintainer resolution appears in the surveyed thread. Between 2026-10-04 and 2026-11-05 many other reporters corroborated the pattern: explicitly authorized GitHub, Slack, Gmail, Drive, Sheets, Calendar and browser actions returning the same user-cancelled message, with each reporter stating they had not cancelled anything, and several noting that a retry with the same arguments later succeeded. One commenter proposed a three-layer split of the failure (authorization delivery, authority consumption, cancellation attribution) and a decision trace to diagnose it. A 2026-10-30 report added a recovery through a saved scoped rule, followed by a recurrence the same day. No maintainer fix or closure decision appears in the surveyed thread. The claim documents only the original report, not a verified current state. Falsification probe: If issue 50769 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn. Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by scripts/verify_claim_verbatim.py.

See also: Codex vs Cursor head-to-head comparison on 9bests