Record typed Persona consequence boundary

This commit is contained in:
Meta Vens 2026-08-23 16:48:31 +02:00
commit 4e4d35ea36
5 changed files with 28 additions and 4 deletions

View file

@ -224,7 +224,7 @@ Delete or demote:
Keep the five-day shakedown, Ox17 deployment, and Model Atlas operational Gate
1 paused. Continue the source-subtraction and test-authority audit from exact
`500125d5`. The current sequence has deleted the semantic-memory/workspace-coverage
`63939fa2`. The current sequence has deleted the semantic-memory/workspace-coverage
projection system, its local daemon supervisor, and the surviving generic
RepoModel profile, impossible lifecycle, producerless receipt/summary, and
decorative scoring axes. Modeling consumes typed Body and keyed RepoModel state
@ -526,6 +526,12 @@ Cut in this order:
projection, rendering, and operator-thought sealing live once in
`epiphany-core`; the coordinator imports that module directly. Cargo falls to
20 executable targets without losing the operator-sealing check.
40. **Persona consequence safety is typed at `63939fa2`:** projector and Persona
prose are inert stage outputs, so a hand-written forbidden-word tribunal
cannot add authority. Only the Interpreter's closed effect enum can propose
consequence. The tribunal, test-only prompt wrapper, prompt-substring test,
and forbidden-word test are deleted; exact stage/context and typed-effect
refusal tests remain.
Do not preserve the aggregate for compatibility, manufacture a bootstrap
thread, release autonomous scheduling, register topology in `gamecult-ops`,