fix(prose-tests): the asserter is told which substitutions were armed

[leeovery]

Summary

• The gap the recording created. The action log shows a stub as the walker writing a report an agent would have produced. An asserter that has never heard of the stub reads that as a missing agent dispatch — root-cause-validation-clean-verdict failed 2/5 on precisely that, having behaved correctly throughout. Armed stubs now reach the asserter with their triggers and descriptions, as a distinct prompt section (P4b).
• An authoring rule from the same round (P4c). A claim about something displayed — "offers the choice and waits" — can never be evidenced by the action log, because emitting text is not a tool call. It rests on the walker's narrative, the one piece of evidence it authors itself. Most display claims have a consequence that can be asserted: the answer consumed, the arm taken, the state produced. Prefer the consequence.
• Deletes smoke-boot-in-world. It absorbed three authoring fixes in one day — a contradictory stop point, then claims about renderings after that stop, then a scope anchor (#Boot) fighting its own claims — for a framework smoke test whose job start-lists-active-work already does properly and passes 5/5.

Test plan

• npm test green: 1698 tests, 0 fail (one fewer suite entry: the deleted case's snapshot check).
• Corpus revalidates: 12 cases, 1 stub.
• root-cause-validation-clean-verdict should now be judged correctly rather than failing on its own substitution — worth a re-run once this lands.

🤖 Generated with Claude Code