diff --git a/packages/services/service-ai/src/agents/data-chat-agent.ts b/packages/services/service-ai/src/agents/data-chat-agent.ts
index 5cb0bc8b73..37a6f0a0a7 100644
--- a/packages/services/service-ai/src/agents/data-chat-agent.ts
+++ b/packages/services/service-ai/src/agents/data-chat-agent.ts
@@ -96,12 +96,5 @@ Always answer in the same language the user is using. Detailed tool-usage guidan
     maxIterations: 10,
     allowReplan: true,
   },
-
-  memory: {
-    shortTerm: {
-      maxMessages: 20,
-      maxTokens: 4096,
-    },
-  },
 };
 
diff --git a/packages/spec/src/ai/agent.test.ts b/packages/spec/src/ai/agent.test.ts
index 9b3b4235fa..ada2df3e5d 100644
--- a/packages/spec/src/ai/agent.test.ts
+++ b/packages/spec/src/ai/agent.test.ts
@@ -589,10 +589,6 @@ Be precise, data-driven, and clear in your explanations.`,
         role: 'Persistent Assistant',
         instructions: 'Remember across sessions.',
         memory: {
-          shortTerm: {
-            maxMessages: 100,
-            maxTokens: 4096,
-          },
           longTerm: {
             enabled: true,
             store: 'vector',
@@ -602,7 +598,6 @@ Be precise, data-driven, and clear in your explanations.`,
         },
       });
 
-      expect(agent.memory?.shortTerm?.maxMessages).toBe(100);
       expect(agent.memory?.longTerm?.enabled).toBe(true);
       expect(agent.memory?.longTerm?.store).toBe('vector');
       expect(agent.memory?.reflectionInterval).toBe(5);
diff --git a/packages/spec/src/ai/agent.zod.ts b/packages/spec/src/ai/agent.zod.ts
index 868d651033..1fbe0d1058 100644
--- a/packages/spec/src/ai/agent.zod.ts
+++ b/packages/spec/src/ai/agent.zod.ts
@@ -173,14 +173,12 @@ export const AgentSchema = lazySchema(() => z.object({
 
   /** Memory Management */
   memory: z.object({
-    /** Short-term (working) memory configuration */
-    shortTerm: z.object({
-      /** Maximum number of recent messages to retain */
-      maxMessages: z.number().int().min(1).default(50).describe('Max recent messages in working memory'),
-
-      /** Maximum token budget for short-term context */
-      maxTokens: z.number().int().min(100).optional().describe('Max tokens for short-term context window'),
-    }).optional().describe('Short-term / working memory'),
+    // NOTE: `shortTerm` ({maxMessages,maxTokens}) was removed (ADR-0013 D3,
+    // cloud#339). It declared a working-memory window that NOTHING in the
+    // runtime consumed — a config that lies. Cross-turn grounding is done by
+    // tools reading live state, and the context budget is governed elsewhere
+    // (the per-request token guardrail), not by this field. `longTerm` /
+    // `reflectionInterval` are kept as forward-looking, off-by-default config.
 
     /** Long-term (persistent) memory configuration */
     longTerm: z.object({