fix(agents): drop the re-wrap funding from the AGENTS.md hunk

REWORK. The previous commit funded ~230 bytes of new practice text by
re-wrapping AGENTS.md §9's 9-line block from ~85-byte to ~115-byte width. That
is 筹行 — re-wrap used to buy lines for content — banned by the maintainer
ruling of 2026-08-17 as sharpened 2026-08-29:

  「筹行(为内容购买行数)⛔ vs 独立密度修复(无内容购买)允许」

The ban binds this file, not only the pm-dispatch corpus. Decisively, the
maintainer ruled TODAY (13:33Z, on #13052) a ceiling raise 1162→1164 to fund
+2 lines in this very §9, whose lines wrap at ~90 bytes; if re-wrap funding
were legal here that ruling would have been pointless. Landing the previous
shape would have overturned that precedent from below.

This commit restores the block to main's byte-exact wrap (blob 8802388) and
keeps exactly one change, edited IN PLACE on its own line with no neighbouring
line re-flowed:

  ⛔ **No hook backs this one** → ⛔ **No hook backs the moving-ref half**

That repair is not optional and is not content purchase: a hook now DOES back
the enumeration half, so the unqualified sentence became false in this PR. The
line goes 87 → 98 bytes against the 120-byte cap, and `git diff` against main
is 1 insertion, 1 deletion, zero net lines — AGENTS.md stays 1162/1162.

The enumerate-from-the-ref idiom is NOT lost. It ships in this PR in the
guard's header and in the guard's BLOCK MESSAGE — where the grading wanted the
explanatory weight, and where an agent meets it at the moment of the mistake —
and in objectui's AGENTS.md, which carries no ratchet. Its objectstack
instruction-text home is deferred to the follow-up card being filed under
option B of this card's open question 1, where the line can be funded legally.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EXxTW8mvPBhoHxmyPZ63de

Author: claude

AGENTS.md

@@ -403,15 +403,15 @@ newer than your base: another agent's merged work entering your tree under the n
 the last fetch in this checkout wins** (⚠️ not per worktree: measured on git 2.43, a
 linked worktree has its own, so a sibling's fetch does **not** move yours and the hazard
 is the **shared primary checkout**, where every agent's first fetch lands before it builds
-one). It fails by **absence**, not wrong content: split from its `git fetch` by any intervening fetch, `git diff
-…FETCH_HEAD` **exits 0 printing nothing** — read as *"the change isn't there"*, a confidently wrong verdict on
-someone else's work, worst for **reviewing** seats. Practices: pin `BASE=$(git rev-parse HEAD)` at worktree
-creation and restore against that commit, never a moving ref name; verify a named remote-tracking ref's content by
-occurrence counts on disk; fetch into a ref you own (`git fetch origin <branch>:refs/<ns>/<id> -f`) and read that;
-diff restored paths against `BASE` before staging; **enumerate from the ref you read** — `git ls-tree --name-only
-origin/main <dir>`, never a working-tree glob feeding per-file `origin/main` reads, whose misses cannot appear in
-its own output (`guard-tree-enum.sh`). ⛔ **No hook backs the moving-ref half** — safe and unsafe spellings are
-both ordinary `git checkout` / `git fetch`, so a mechanical block would fire on correct usage.
+one). It fails by **absence**, not wrong content: split from its `git fetch` by any
+intervening fetch, `git diff …FETCH_HEAD` **exits 0 printing nothing** — read as *"the
+change isn't there"*, a confidently wrong verdict on someone else's work, worst for
+**reviewing** seats. Practices: pin `BASE=$(git rev-parse HEAD)` at worktree creation and
+restore against that commit, never a moving ref name; verify a named remote-tracking
+ref's content by occurrence counts on disk; fetch into a ref you own (`git fetch origin
+<branch>:refs/<ns>/<id> -f`) and read that; diff restored paths against `BASE` before
+staging. ⛔ **No hook backs the moving-ref half** — safe and unsafe spellings are both ordinary
+`git checkout` / `git fetch`, so a mechanical block would fire on correct usage.
 
 **⛔ Nor the build cache** — turbo resolves the repo root through the **common** dir, so every
 worktree replays ONE `.turbo/cache`. Symptom: a typecheck failing on a package your diff never