Add routing guide and subagent thinking overrides - #405
Conversation
Implementation Plan — Stage 1Problem and approachStage 1 has two focused deliverables:
Per the user mandate, guide generation stays entirely in External prior art supports this narrow design:
Deliverables1. Built-in
|
Progress UpdateImplemented Stage 1:
Verification:
Commit: Progress tracked by mach6 |
Vitest coverage
|
Code ReviewCriticalNone. ImportantFinding 1 — Fallback-selected model capability validation lacks an integration testSeverity: high · Confidence: 97
SuggestionsFinding 2 — Effective-thinking telemetry test does not distinguish child-reported metadata from the requestSeverity: medium · Confidence: 98 The explicit-level test requests Finding 3 — Continued-loop and default-
|
Review AssessmentClassifications
Action Plan
Assessment by mach6 |
Progress UpdateFixed genuine review findings 1–6:
Verification:
Commit: Progress tracked by mach6 |
Closes #394
Stage 1 adds the model-routing-guide research skill and explicit per-subagent thinking overrides with effective thinking observability.
Implementation plan posted as a comment below.