Bump max_tokens to 1536 to avoid truncation on long assembly

mattgodbolt-molty · mattgodbolt-molty · commit 508c2a2dbe30 · 2026-02-21T11:50:23.000-06:00
Two test cases were hitting the 1024 limit and getting cut off
mid-explanation. 1536 gives enough headroom without encouraging
verbosity.

🤖 Generated by LLM (Claude, via OpenClaw)
diff --git a/app/prompt.yaml b/app/prompt.yaml
@@ -2,7 +2,7 @@ name: Production Sonnet 4.6
 description: Sonnet 4.6 with deduplicated, tighter prompts
 model:
   name: claude-sonnet-4-6
-  max_tokens: 1024
+  max_tokens: 1536
   temperature: 0.0
 audience_levels:
   beginner:
