fix(prose-tests): prescribe the cross-check, drop a check that cannot fire, lint the scope

[leeovery]

Three things the full run turned up, none of them failures.

The asserter's cross-check is now prescribed

It had begun comparing the action record against the walker's narrative unprompted, and reporting divergences. Right instinct — but nobody's instruction, so any run depending on it was depending on a coincidence.

Now required, both directions named:

• Narrated but not recorded — the walker described something it didn't do. The step resting on that claim fails, and the discrepancy is a finding in itself.
• Recorded but not narrated — usually orientation, fails nothing, but reported: a call that touched state and went unmentioned is the shape a silent repair takes.

The two streams exist so neither is taken on trust. That only holds if they're actually set against each other every time.

engine_before_write declared only where it can fire

It was on all ten cases and could only ever trip on the three that write state. Reporting N/A seven times is honest, but it's still a check dressed as coverage.

The check itself stays — it's the anti-fabrication guard for anything landing state, and state-landing cases are the direction we're heading.

New: the scope lint

A case's files[] isn't documentation — it decides whether editing a file selects that case for a run. Prose the walk goes through but the case omits can be changed without the test covering it ever running again. Every case has such gaps today.

It can't be built by following links: the reachable set is far wider than the walked one, and a case declaring everything reachable would be selected for branches it deliberately stops before (session-loop.md, task-loop.md, conclude-discovery.md…). So it's computed from the walk's own Read calls, and reported as a finding about the case — correcting nothing, since which files belong is an authoring judgement.

Test plan

• 86/86 across the prose suites, including five new cases for the lint: it names undeclared prose, stays quiet when the list is complete, ignores world-state reads and shell cats, and reports each file once
• Corpus valid

🤖 Generated with Claude Code

Stack

1. docs(design): prose-tests programme design log #544
2. feat(prose-tests): the framework — cases, worlds, runner, skill #545
3. test(prose): feature happy-path corpus — five worlds, seven cases #546
4. test(prose): bugfix corpus — the investigation-centric surfaces #548
5. test: retry recursive teardown removals — kill a class of phantom failures #549
6. fix(entry-skills): close the handoff fences — six files render their arms wrong #550
7. docs: a contributing page for working on the system #551
8. fix(entry-skills): every handoff arm says to invoke the skill #552
9. fix(implementation): environment setup belongs to the setup reference alone #553
10. fix(prose-tests): the asserter is told which substitutions were armed #554
11. feat(prose-tests): the mid-flow substitution, and a world only prose can describe #555
12. test(prose): claims assert consequences, not what was displayed #556
13. feat(prose-tests): record everything the agents do, results included #557
14. fix(discussion-entry): the handoff reports the source it actually had #558
15. fix(prose-tests): the stop hook records, and names the model that walked #559
16. fix(prose-tests): command output was never actually recorded #560
17. feat(prose-tests): judge the walk as told, not the summary returned #561
18. feat(prose-tests): decide in code what an agent should not be deciding #562
19. test(prose): a case starts where a session starts #563
20. feat(prose-tests): walk on Sonnet, judge on Opus, escalate a failure #564
21. test(prose): give the eight read-only cases something that can fail #565
22. test(prose): only walks that can be observed, and checks that survive the trip #566
23. fix(prose-tests): the verdict names only the model the record names #567
24. test(prose): discovery, walked to the point where work first exists #568
25. fix(prose-tests): the asserter judges which of prose or walker was at fault #569
26. docs(conventions): a step whose reference routes every exit still signposts #570
27. test(prose): discovery's epic arm, to the same durability boundary #571
28. fix(prose-tests): keep the walk's closing turn, and name the one live difference #572
29. fix(prose-tests): prescribe the cross-check, drop a check that cannot fire, lint the scope #573 👈 current
30. test(prose): declare the prose each case actually walks #574
31. fix(investigation): the symptom interview never ran for work shaped in discovery #575
32. feat(prose-tests): conduct — playing a person where a script cannot reach #576
33. fix(prose-tests): what a write put in a file is evidence, not incident #577
34. fix(investigation): the interview starts from the carrier, not over it #578
35. test(prose): quick-fix scoping, the whole definition stage in one walk #579
36. fix(prose-tests): the world log is the walker's record #580
37. fix(prose-tests): a confirmed failure reports both runs' checks #581
38. fix(prose-tests): each prescribed command runs as written #582
39. fix(scoping): context gathering reads the whole carrier, asks only the gaps #583