From 12f283e7c9e52b341aed9fd12c61de464f01cc72 Mon Sep 17 00:00:00 2001
From: David Ibia <github@boxpositron.dev>
Date: Sat, 10 Oct 2026 14:56:06 +0100
Subject: [PATCH] feat(omo): route impeccable critique to dedicated model
 categories

Impeccable's critique runs two isolated assessments. design-critique-a
runs the design-director review on Claude Opus 5.5 at xhigh, with Fable
5.1 behind it; Opus 5.5 has the top Design Arena website score of the
models in this config. design-critique-b runs the detector and browser
pass on GPT-6.1 Sol at high, with Sonnet 5.5 behind it so the second
assessment still runs without a ChatGPT login. critique.md sends each
assessment to its category.
---
 .config/omo/omo.jsonc                         | 27 +++++++++++++++++++
 .../skills/impeccable/reference/critique.md   |  2 ++
 2 files changed, 29 insertions(+)

diff --git a/.config/omo/omo.jsonc b/.config/omo/omo.jsonc
index 4c1c438..ef7a3af 100644
--- a/.config/omo/omo.jsonc
+++ b/.config/omo/omo.jsonc
@@ -110,6 +110,33 @@
         "reasoning": "high",
         "prompt_append": "Before substantive UI work, load frontend-workflow. Follow the product's design context, keep one writer per file set, inspect actual desktop and mobile screenshots, exercise the primary interaction, and report verification evidence. Stop when acceptance criteria pass."
       },
+      // impeccable critique. A = design-director review (taste, heuristics): Opus 5.5 has the top Design Arena
+      // website Elo of the models we run (1339 vs Sonnet 5.5 1322, Fable 5.1 1318; GPT-6 unranked, Oct 2026).
+      // B = detector + browser evidence (mechanical tool use): GPT-6.1 Sol, with a Claude fallback so B stays available.
+      "design-critique-a": {
+        "models": [
+          {
+            "model": "anthropic-subscription/claude-opus-5-5",
+            "reasoning": "xhigh"
+          },
+          {
+            "model": "anthropic-subscription/claude-fable-5-1",
+            "reasoning": "xhigh"
+          }
+        ]
+      },
+      "design-critique-b": {
+        "models": [
+          {
+            "model": "chatgpt-subscription/gpt-6.1-sol",
+            "reasoning": "high"
+          },
+          {
+            "model": "anthropic-subscription/claude-sonnet-5-5",
+            "reasoning": "high"
+          }
+        ]
+      },
       "writing": {
         "models": [
           {
diff --git a/.config/opencode/skills/impeccable/reference/critique.md b/.config/opencode/skills/impeccable/reference/critique.md
index d3a9362..dedf037 100644
--- a/.config/opencode/skills/impeccable/reference/critique.md
+++ b/.config/opencode/skills/impeccable/reference/critique.md
@@ -32,6 +32,8 @@ Resolve one stable target, run two independent assessments, synthesize a design
 
 Delegate Assessment A and Assessment B to separate sub-agents. They must not see each other's output. Do not show findings to the user until synthesis.
 
+OmO routing: when the task tool accepts a category, spawn Assessment A with `category: "design-critique-a"` and Assessment B with `category: "design-critique-b"`. Never pass `model` with a category.
+
 Sub-agent gate (all harnesses):
 - Unless a harness-specific gate below overrides this, spawn A and B as two isolated, parallel sub-agents whenever a sub-agent/Task tool is exposed. This is the default and is mandatory; do not run them inline because it is faster.
 - "Unavailable" means exactly one thing: no sub-agent/Task tool is exposed in this session (or, on harnesses that ask, the user declined). It does not mean inconvenient.