fix(delegate-task): require one goal per deep call

The caller-facing description told orchestrators WHAT `deep` is but
never specified goal cardinality, so they bundled multiple goals into
one call (e.g. "fix X + merge Y + deploy Z"). The deep agent's own
prompt was already configured to refuse such bundles, but only after
they arrived. The rule now lives on the caller side: ONE goal + ONE
deliverable per call; multiple goals must fan out as parallel `deep`
calls.

src/tools/delegate-task/openai-categories.ts

@@ -203,7 +203,7 @@ export const OPENAI_CATEGORIES: BuiltinCategoryDefinition[] = [
   {
     name: "deep",
     config: { model: "openai/gpt-5.5", variant: "medium" },
-    description: "Goal-oriented autonomous problem-solving. Thorough research before action. For hairy problems requiring deep understanding.",
+    description: "Goal-oriented autonomous problem-solving on hairy problems requiring deep research. ONE goal + ONE deliverable per call — multiple goals must fan out as parallel `deep` calls, never bundled into one.",
     promptAppend: DEEP_CATEGORY_PROMPT_APPEND,
     resolvePromptAppend: resolveDeepCategoryPromptAppend,
   },