Record the whole-manual warmth sweep and its rulings in the handoff

[dcellison]

The handoff package catches up with the manual warmth sweep (#464 through #472), following the pattern #463 set for the Part I arc. Three files change; the package's counts, sequences, and other sections were checked against the live repository and remain accurate.

What the refresh records

• README.md, where the work stopped: the Part I warmth sentence becomes the whole-sweep record, naming the method (hedge rationing, imagery on design-intent verbs, part thresholds), the additive make-up pass over the middle grammar chapters, and the two rules the voice guide gained (the waiver row with its ceiling reading, the effect-verb-under-an-image row with its gallery pair).
• README.md, the maintainer's working contract: one new standing instruction from a direct correction. When a commission names an additive goal, track the additions per pass rather than only the corrections; declining effort across an arc is a finding to raise before Daniel has to raise it.
• current_state.md: the snapshot header and the warmth paragraph now describe the completed sweep, including the ceiling ruling in operational terms (a kept statement teaches its own chapter's subject; the general charter lives once, in Part I; a chapter may carry none) and the closing instruction to measure additions on additive commissions.
• language_and_voice.md: the voice-prohibitions section gains the two rules in summary form so a replacement model meets them before reaching the full guide.

What was checked and left alone

The execution sequence (pamphlets/complementizers next, then the remaining pamphlets and current documents, then book chapter 8) is unchanged and still correct. The lexicon counts, decision-register table, module state, corpus state, and book state sections were verified against the live repository and need no update; no vocabulary changed during the sweep. repository_workflow.md already documents the phonetic-neighbor CI check that briefly caught #469, so no addition was needed there; the failure was a reading gap, not a documentation gap.

Audit

Full protocol on the new prose: all hard-zero rows at zero, no Phi citations added, single-line paragraphs throughout. Validator: 1275 entries, 0 errors, 0 warnings.