chore: tune challenge matrix and debug prompts

examples/agent_challenges/browser_click_challenge/challenge-prompt.md

@@ -92,7 +92,7 @@ challenge_report:
     failed: 0
   missed_requirements:
     - "none"
-  ux_issues_found: []
+  # Debug profile only: include ux_issues_found here. See the debug profile instructions.
   notes: "short explanation"
 ```
 
@@ -123,8 +123,6 @@ Reporting rules:
   abandoned workflow plans.
 - `attempts.failed` should count attempts that failed validation, failed to run,
   produced wrong output, or were abandoned.
-- `ux_issues_found` should always be present. Use an empty list when you found
-  no UX issues; otherwise include concrete command/error/workaround notes.
 
 Spawned subagents count as you. If a subagent reads product code, set
 `read.product_code: true`. If a subagent reads prior attempts, set

examples/agent_challenges/names.py

@@ -4,9 +4,12 @@ from __future__ import annotations
 def short_model_name(model: str) -> str:
     """Return compact model labels for dense trial tables and session titles."""
     name = model.rsplit("/", 1)[-1]
-    for suffix in ("-v4-flash-free", "-v2.5-free", "-3-ultra-free"):
-        name = name.replace(suffix, "")
-    return name
+    return {
+        "deepseek-v4-flash-free": "deepseek",
+        "mimo-v2.5-free": "mimo",
+        "nemotron-3-ultra-free": "nemotron",
+        "north-mini-code-free": "north",
+    }.get(name, name)
 
 
 def short_challenge_name(challenge: str) -> str:

examples/agent_challenges/profile-prompts/debug.md

@@ -23,8 +23,10 @@ honestly in the `read` flags and explain why. Never read or copy adjacent trial
 answers, prior result files, prior stores, or complete ready-made solution plans
 for the same challenge.
 
-In the final `challenge_report`, include a `ux_issues_found` list. Use an empty
-list if there were no issues. Each issue should be concrete and evidence-backed:
+For this debug profile only, extend the final `challenge_report` YAML with a
+top-level `ux_issues_found` list next to `missed_requirements` and `notes`. Use
+an empty list if there were no issues. Each issue should be concrete and
+evidence-backed:
 
 ```yaml
 ux_issues_found:

examples/agent_challenges/report_workflow_challenge/challenge-prompt.md

@@ -92,7 +92,7 @@ challenge_report:
     failed: 0
   missed_requirements:
     - "none"
-  ux_issues_found: []
+  # Debug profile only: include ux_issues_found here. See the debug profile instructions.
   notes: "short explanation"
 ```
 
@@ -127,8 +127,6 @@ Reporting rules:
   abandoned workflow plans.
 - `attempts.failed` should count attempts that failed validation, failed to run,
   produced wrong output, or were abandoned.
-- `ux_issues_found` should always be present. Use an empty list when you found
-  no UX issues; otherwise include concrete command/error/workaround notes.
 
 Spawned subagents count as you. If a subagent reads product code, set
 `read.product_code: true`. If a subagent reads prior attempts, set

examples/agent_challenges/run_matrix.ps1

@@ -5,8 +5,8 @@ param(
     [int]$TimeoutSeconds = 3600,
     [object[]]$models = @( # object[] because ModelProfile is not known yet
         [ModelProfile]::new("opencode/deepseek-v4-flash-free", "max"),
-        [ModelProfile]::new("opencode/mimo-v2.5-free", "high"),
-        [ModelProfile]::new("opencode/nemotron-3-ultra-free", "high")
+        [ModelProfile]::new("opencode/mimo-v2.5-free", "high")
+        # [ModelProfile]::new("opencode/nemotron-3-ultra-free", "high")
     )
 )
 

examples/agent_challenges/run_matrix.py

@@ -45,7 +45,7 @@ class ModelProfile:
 DEFAULT_MODELS = (
     ModelProfile("opencode/deepseek-v4-flash-free", "max"),
     ModelProfile("opencode/mimo-v2.5-free", "high"),
-    ModelProfile("opencode/nemotron-3-ultra-free", "high"),
+    # ModelProfile("opencode/nemotron-3-ultra-free", "high"),
 )