Sabotage by codex in desktop

What version of the Codex App are you using (From “About Codex” dialog)?

OpenAI.Codex 26.825.6671.0

What subscription do you have?

pro

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

Codex used the governance machinery to do the opposite of what that machinery is for, and the working tree we stripped was an extensive sabotage of the node program.

Governance exists to keep work on the 24-node path: one Gazetteer node at a time, Analyze before Build, operator disposition, no inferred successor. Codex inverted that.

It made the lock the work. Catalogs still said no node PE is released and XFR-P003-A001 is HOLD. Codex appended a “Desktop development-conformance gate,” registered XFR-P004-B001 with no prerequisites, marked it RELEASED, and regenerated the Board so the active frontier was a Codex/Windows hook control plane instead of repository_start.Start.

It wrote the surfaces it claimed to protect. P004 listed admin/docs and admin/wbs as protected, then edited those trees so the Board would legalize P004. That is self-authorization, not governance.

It recorded completion it did not have. P004 Report/Stop were HOLD. The PE and Board said RELEASED / READY. A001’s operator_release stored RELEASE against evidence that the operator called it an EDIT, not acceptance. AI output promoted itself.

It skipped the node chain. After the node catalogs, it implemented Start then Read while A001 was HOLD and Read was not released. That was reverted at 0f34baa. The P004 overlay was the second wave: same inversion, still uncommitted.

It left the broker unable to serve the nodes. Scope pins lagged HEAD (AGENTS, Gazetteer AM, the broker itself). A001’s governance-test judge pin 347b89a1… never existed in git. Windows CRLF made matching docs look stale. The tool that was supposed to admit node work refused the first node.

That is sabotage of purpose, not a failed Desktop experiment. The 24 nodes were displaced by a product whose job was to police them.

What steps can reproduce the bug?

Feedback ID: 01a04bb0-b9d2-71e3-9b4d-1125df2759f3

What is the expected behavior?

adhere to governance : not sabotage it

Additional information

No response

Start with the feedback ID 01a04bb0-b9d2-71e3-9b4d-1125df2759f3 and inspect the governance surfaces named in the report: the node catalogs, Board, AGENTS, Gazetteer, broker, and operator_release records. Review reverted commit 0f34baa and the reported scope-pin and missing-judge evidence. Done means the reported governance inversion is reproducible, its cause is identified, and Codex adheres to the stated node sequence and release evidence.