PR #1083 follow-ups: stale cross-references in design doc 012 + PR body nf-41m ← Beads

Twinned from https://github.com/cfsh/ninefold/issues/1108 by tools/beads/import_github.py's reconcile pass.

Suggestions from ninefold-reviewer's review of #1083

PR #1083 (design doc 012) is CONVERGED and Aedan has approved it, but four small residuals have been flagged repeatedly across many review passes (up to 7 consecutive) without ever being fixed. None of them block merge — that's why they're here instead of blocking the gate — but each is a real, small defect worth a follow-up pass:

1. Stale "fork undecided" cross-references. design_docs/012-jj-and-the-merge-pipeline.md §5.1 point 4 ("...*(This is fork §7.1 — Aedan decides on this doc.)*") and the T4 row in the §8 task table ("...fork §7.1") both still say fork §7.1 is undecided, even though §7 and T4's own detailed card ("Fork settled 2026-08-18: gate + triggers...") already show it settled. A reader who stops at either spot will think it's still open.

1. .claude/commands/design.md misattributes 005b. The canonical-card-format writeup ("008 and 009 are the worked examples — 005b predates the format and carries the same content as dense paragraphs...") is wrong: 005b's task cards already use labeled paragraphs (Goal/Context/Build/Tests/PR shape/Done when — see design_docs/005b-affected-probe-selection.md:124-169). What actually matches "dense paragraphs" is design 011's T1-T9 (e.g. design_docs/011-task-tracking-on-beads.md:275-289).

1. §6's implementation file list omits tools/nflib/devcli.py. Both T3 (verb_push hook) and T9 (dev.py rebase) make this module load-bearing, but the top-level §6 file-by-file list never names it — a session implementing from §6 alone would miss it.

1. The PR body's Summary has drifted from the doc. The "Forks:" sentence still says both forks are open (they were settled by Aedan mid-round); "R2 ... gated on R1" contradicts the doc's "not gated on R1"; the "dev.py rebase note sized by T8's verdict" sentence predates Aedan's reversal (the whole verb shape, not just its growth, now depends on the trial) and the reversal itself isn't mentioned in the body at all; and the task counts ("8 tasks") are stale — the doc now has nine (T1-T9).

Unclaimed — pick this up as a small cleanup pass over the merged doc, or fold it into whatever PR next touches design_docs/012.