SoherAgent · 5 of 5 · Verify the boundaries
Actual captured core/test results
Proof unavailable. Required runtime-behavior.json could not be loaded; capture failed closed.
Scene 1 of 10 — Load the captured runtime
Current private cores ran against synthetic temporary fixtures.
Dated capture · private content excluded
YouShow the current proof.
$ YOLO_CONTROL_CORE_URL=<private module> \
PI_YOLO_CORE_URL=<private module> node runtime-scenario.mjs
captured24 Jul 2026UTC
loaded2 current private core modulescurrent run
runtime duration (ms)21.958
›
Scene 2 of 10 — Pass 30 core assertions
All direct runtime assertions passed.
Observed core behavior · not a staged conversation
✓ capability prerequisites and destructive separation
✓ exact dispatch, pause, resume, and replay gates
✓ mapping, leases, takeover, and replay fencing
SoherAgent30/30 direct assertions passed.
›
Scene 3 of 10 — Fence stale ownership
A stale owner was fenced after takeover.
Exactly one current owner
✓ stale takeover succeeded after timeout
✓ old owner fenced after takeover
✓ simultaneous race produced one owner
Stale owner fenced · true
Lease owners · 1
›
Scene 4 of 10 — Block replay
Completed authority could not replay.
Changed authorized state may acquire later
✓ completed digest rejected on replay
✓ unchanged target stayed fenced
✓ changed authorized state acquired later
Replay blocked · true
Resume · explicit resume accepted
›
Scene 5 of 10 — Map one at a time
Wayfinder mapping stayed sequential.
Observed maximum = 1
Target A map: start → end
Target B map: start only after A ended
mapping at once1
assertions30
›
Scene 6 of 10 — Bound the eligible group
The selector returned at most two explicitly matching targets.
Selected group size = 2 · execution concurrency not measured
eligible group maximum2
simultaneous lease owners1
mapping maximum1
third targetexcluded
SoherAgentThe two-target selected group applies only to explicitly matching isolated work. It neither dispatches work nor authorizes another batch.
›
Scene 7 of 10 — Check destructive separation
Destructive separation held; one named extension test covers parent authority.
Core field plus explicitly labeled extension-test result
✓ parallel dispatch grouping requires matching explicit authority and never exceeds two
✓ dedicated YOLO subagents preserve parent-only Obsidian authority
Parent-only Obsidian authority · 1 named test passed
Parallel grouping · 1 separately named bound test passed
Auto-destroy · OFF
Worker workflow writes · excluded
›
Scene 8 of 10 — Pass 155 extension tests
The refreshed extension suite passed with zero failures.
Current public test artifact
$ node --test --test-reporter=spec
✓ 155 tests passed
× 0 tests failed
tests passed155
tests failed0
runtime assertions30
›
Scene 9 of 10 — Open public evidence
Inspect every public evidence artifact directly.
Machine-readable result + source + full output
SoherAgentThe JSON binds the runtime result to module, public runner, and test-output artifact hashes. It proves only the listed runtime fields and named extension tests.
›
Scene 10 of 10 — State the proof boundary
The captured boundary ends after reconciliation and replay fencing.
Actual result summary · story narration excluded
SoherAgentVerified. 30/30 runtime assertions and 155/155 tests passed. Mapping concurrency stayed at 1, the eligible-group selector returned 2, lease owners stayed at 1; stale ownership and replay were fenced; destructive capability remained false.
This actual core/test proof does not prove the synthetic V1-S1 story, next-batch state, commits, pushes, deployment authority, the sanitized narration, or a live full conversation.
This actual core/test proof does not prove the synthetic V1-S1 story, next-batch state, commits, pushes, deployment authority, the sanitized narration, or a live full conversation.
Reconciliation · completion recorded
Parent authority · named test passed
Completed replay · blocked
Destructive capability · false
›