fix: cap AI tool result size to prevent JSON parse errors

When generating multiple cover images in one conversation, the
accumulated tool results (each ~400KB base64) caused the OpenAI
conversation payload to exceed JSON parsing limits in the browser.

Fix:
- Strip coverImageUrl from invalidate action results (not needed by AI)
- Cap invalidate results to 4KB
- Cap all tool results to 8KB
- Prevents "JSON.parse: unexpected character" errors during batch ops

Co-Authored-By: claude-flow <ruv@ruv.net>

packages/api/src/router/assistant-tools.ts

@@ -5564,16 +5564,19 @@ export async function executeTool(
 
       if (actionType === "invalidate") {
         const scope = actionResult.scope as string[];
-        // Strip __action and scope from the result sent back to the AI
-        const { __action: _, scope: _s, ...rest } = actionResult;
+        // Strip __action, scope, and large data from the result sent back to the AI
+        const { __action: _, scope: _s, coverImageUrl: _img, ...rest } = actionResult;
+        const content = JSON.stringify(rest);
         return {
-          content: JSON.stringify(rest),
+          content: content.length > 4000 ? content.slice(0, 4000) + '..."' : content,
           action: { type: "invalidate", scope },
         };
       }
     }
 
-    return { content: JSON.stringify(result) };
+    // Cap tool result size to prevent oversized OpenAI conversation payloads
+    const content = typeof result === "string" ? result : JSON.stringify(result);
+    return { content: content.length > 8000 ? content.slice(0, 8000) + '..."' : content };
   } catch (err) {
     const msg = err instanceof Error ? err.message : String(err);
     return { content: JSON.stringify({ error: msg }) };