Description
SKT-0011 scenarios C5, D5, D6, and D8 could reach the control surfaces, but the UI path yielded only a redirect in the clean-container check, inventory exposed no per-blueprint entries, diagnostics did not establish shell-free sufficiency, and the full catalogue exposed zero scalable targets. The round could not verify every UI view, identity separation, or a real scaling mutation from the supplied control data.
Acceptance Criteria
- #1 The admin UI has an executable check that enumerates every shipped view and confirms it reflects live state
- #2 Control inventory exposes per-blueprint and substrate identity needed for label-separation diagnosis
- #3 Diagnostics identify a failing blueprint and lane without container shell access
- #4 At least one shipped acceptance configuration exposes a valid scalable target, or the API clearly reports why none exist
- #5 Reset and scaling acceptance checks assert an observable emitted-data change
Definition of Done
- #1 just check (fmt-check, lint, gen-check, env-check, docs-check, test, race, hygiene, ui-check, compose-check, helm-test, lab-check, signal-fidelity)
- #2 just gen (only if a blueprint field, construct/workload config struct, or a skill under plugins/synthkit/skills/ changed)
- #3 just dump — inventory diffed against signals/
Implementation Plan
Lane B makes UI, inventory, diagnostics, reset, and scaling self-verifying; root performs emitted-data mutation checks live.
2026-08-31 Root-owned: add a failing runner regression for the inert scale multiplier, inspect the prior high_dpm floor seam, fix composition-root propagation, and prove both scale and reset transitions in emitted live rate without changing the assertion.
Implementation Notes
2026-08-30 closeout: UI enumeration, per-blueprint/substrate inventory, lane diagnostics, and a scalable acceptance target landed. Live scale 2 to 4 and reset were accepted and read back, but emitted rate ratios were 0.977853 while scaled and 0.995540 after reset.
2026-08-31 live proof: scaling the declared application service from 2 to 4 changed the two-minute server-span increase from 926.04 to 2120.12, a 2.29x ratio. Reset produced 966.01, 1.04x baseline, and cleared the control scaling map. HTTP status alone was not used as evidence.
Final Summary
Parked at AC#5. Resume in the runner scaling path and require the selected target live rate to move with the configured multiplier before accepting reset/scaling as self-verifying.
2026-08-31: Propagated the scale multiplier through the runner to the declared application service and proved scale plus reset through emitted live span-rate movement. The final gates passed.