Everything I could decide myself is decided. What's left needs
you. Each one has a recommendation from Fable — if you agree, pick it and
press Done. "I'm not sure" is a real answer and becomes work for me.
You don't need to open the session — pressing Done sends your answer back and it picks up.
1The review-gate fix (PR #61) is finished and correct, but it cannot merge by any path. Landing it needs a deliberate, temporary change to main's branch protection. Which way do you want it done?
Fable recommends: Briefly turn OFF 'include administrators', merge #61, turn it straight back on It is the narrower of the two openings. Turning off include-administrators lets ONE named person deliberately merge ONE pull request; every other open PR stays gated exactly as it is. Removing 'review' from the required checks instead un-gates EVERY open pull request for the whole window, so anything with auto-merge on could land unreviewed while the door is open and nobody would necessarily notice. Both are reversible in seconds; only one of them can let something through by accident. Background you need: the fix itself is not a loosening — I checked the verdict-reading code byte for byte against main and it is identical, 2280 bytes both sides. The reason #61 cannot merge is that two correct safety features meet: the review robot refuses to run on any PR that edits its own workflow file (an anti-tampering rule worth keeping), so the review check fails honestly, and 'include administrators' means nobody can override it. That is permanent and applies to every future PR that touches this file, so it is a decision worth making once, on purpose. Fable was asked, and escalated. Receipt f075528e9 — NEEDS HUMAN. THE TWO PASSES DISAGREED, AND ONLY THE OPTION ORDER DIFFERED. as-given → "Briefly turn OFF 'include administrators', merge #61, turn it straight back on" (high) reversed → "Briefly remove 'review' from the required checks, merge #61, put it straight back" (medium) An answer a reordering can flip is not an independent judgement, it is position bias wearing one. Escalating with both readings attached.
as-given reasoning: Read PR #61 and review.yml. The fix is real (channel bug: track_progress forces the MCP comment tool whose sanitizeContent strips the HTML marker; encoder byte-identical… Tried first, unsuccessfully: Receipt f075528e9. Fable was asked twice with the options in opposite order and genuinely split: asked one way it chose turn-off-include-administrators (high confidence), asked the other way it chose remove-the-review-check (medium). An answer that flips when you reorder the options is not an independent judgement, so it escalated rather than pretending to a verdict. What BOTH passes agreed on, and what I therefore treat as settled: leaving #61 unmerged is not viable. The deadlock is permanent, it currently gates a security fix you already ordered merged on 2026-08-10 (#51), and there is no third mechanism — required checks block direct pushes too, and hand-writing the status would be impersonation since the CLI authenticates as you. What Fable could not settle, and why this is yours: whether opening protection on main is acceptable at all, which of the two openings you prefer, and who performs it.