[Grok] Optimal ACM dao-shu guidance (refined from #4+#5)

[KorenKrita]

Summary

Grok’s optimal always-on ACM guidance: take #5’s de-duplicated spine and fold in the best of #4.

Kept from #5

• Single Recognizable moments table (reasonable, not required) — no Judgment-triggers duplicate
• NEXT stalls only under Failure shapes (not in smells)
• Nudge ↔ CORE vocabulary alignment
• Hard mixed-batch note at runtime/handoff layer
• CORE < 8500 chars

Taken from #4

• Causal opening (process accumulates → ACM keeps growth recoverable)
• Softer fold/rebase criteria naming (vs gate)
• Smells framed against the working-set invariant
• Travel-task cue: handoff working set

Structure

Working-set invariant
Vocabulary
ACM preflight (cost curve + managed chain)
Decision smells (4; no post-fold stalls)
Recognizable moments
Fold criteria + one-line self-check
Rebase criteria
Failure shapes
Handoff + Why examples
Skill pointer

Intended to supersede #2–#5 once reviewed.

Test plan

• bun run verify:acm
• Live spot-check: smell → criteria self-check → travel; pressure nudge urges rebase check

[KorenKrita]

Superseded by #12 (chosen merge target for ACM dao-shu guidance). Closing this alternative.