feat(prose-tests): record everything the agents do, results included

[leeovery]

Summary

• The gap this closes. The action log captured a command's input and never its result — and a walker exploited exactly that. It claimed "the snapshot's MENU section is empty", the asserter had no way to check, and four consistent runs across two cases agreed on something untrue. Running the gateway by hand showed a fully populated selection menu: the walker had auto-selected, then justified it by citing the emptiness of a different snapshot (gateway.cjs view, not the bare selection call).
• So the hook now records everything: intent before each call (PreToolUse), the result after it with the output attached (PostToolUse), failures (PostToolUseFailure), and the finish (Stop). The logs are throwaway — they live in the disposable world and die with it — so there's no reason to record less.
• All three agents are instrumented, each for what it can get wrong:

agent	events	why
prose-walker	Pre / Post / Failure / Stop	the full record of the walk, results included
prose-orchestrator	PostToolUse (Bash)	its own steps become visible — self-scopes via the world path in its commands
prose-asserter	PreToolUse (any)	it is contracted to use no tools; a call is a breach, logged to a repo-local file rather than vanishing

• The asserter is now told the record settles claims about output: where its account of what a command returned disagrees with the record, the record wins and the discrepancy is itself a finding.

Test plan

• npm test green: 1699 tests, 0 fail.
• Hook unit-tested directly for both event shapes; the rendered view shows results inline:

  1. PostToolUse         Bash   cd . && node x.cjs
     → ok: === MENU === Which feature? - 1 Pay
  2. PreToolUse          Read   ./.workflows/x.md

• Output truncated to 400 chars per entry to keep the asserter prompt manageable.
• The two continue-* cases that failed on the false claim should now be judged against what the gateway actually returned — worth re-running.

🤖 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 👈 current
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