Summary

Discovery's Step 5 loads confirm-trigger.md and ends there. Every other step in that skill says where it goes next — so a reader reaching the blank line has one signal left, the next heading, and takes it. A walk did exactly that: read confirm-trigger's route to Step 7, talked itself into Step 6 because the file numbers it earlier, then caught itself and corrected.

The convention already anticipates most of this. A reference whose paths need different backbone steps overrides its caller, because "the caller's single → Proceed to line can only go one place, so the file overrides it." That wording assumes a line exists to be overridden — and for final-review.md it does: three of its eight exits return bare, and the caller's footer is what serves them.

confirm-trigger.md has no such exit. Every path routes, so any footer would name a step that never runs, and the step was left with nothing at all. That's the one shape with no documented form.

Two rules added under ### Load Directive Format:

• Where every exit routes, the footer defers rather than naming a step: → On return, proceed as the reference directed.
• A footer governs only the reference's bare → Return to caller. exits — never branches the reference routed itself. True already, unwritten, and what lets a reader over-apply a footer to all eight of final-review's exits when it serves three.
• And: a step is never left with no footer. Silence reads as the end of the step, making the heading below the fall-through by default.

Applied to workflow-discovery Step 5. Behaviour is unchanged — the routing was always complete for anyone who read the reference; this makes it complete for someone who has just finished reading it.

Swept for other instances

Only two other footerless load-steps exist:

• workflow-specification-entry Step 4 — terminal (invokes the processing skill), correct as-is.
• workflow-bridge Step 2 — a different shape: six conditional branches, each a bare Load, none carrying routing. Reads against the every-branch-routes rule, though each continuation ends by invoking a skill, which is terminal and may be legitimately exempt. Surfaced, not touched.

Test plan

• Conventions lint 29/29; prose gate 47/47
• No simulation change: this adds a routing signpost, not an engine verb or a changed call sequence

🤖 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 👈 current
27. test(prose): discovery's epic arm, to the same durability boundary #571