diff --git a/packages/coding-agent/CHANGELOG.md b/packages/coding-agent/CHANGELOG.md
index 03973318..18385c77 100644
--- a/packages/coding-agent/CHANGELOG.md
+++ b/packages/coding-agent/CHANGELOG.md
@@ -8,6 +8,7 @@
 
 ### Fixed
 
+- Fixed compaction and branch-summary requests to use fresh routing session IDs with prompt caching disabled where supported.
 - Fixed explicit self-updates when `PI_SKIP_VERSION_CHECK` is set ([#6977](https://github.com/earendil-works/pi/issues/6977)).
 
 ## [0.81.1] - 2026-07-21
diff --git a/packages/coding-agent/src/core/compaction/compaction.ts b/packages/coding-agent/src/core/compaction/compaction.ts
index 0ae6deeb..3f9b90ca 100644
--- a/packages/coding-agent/src/core/compaction/compaction.ts
+++ b/packages/coding-agent/src/core/compaction/compaction.ts
@@ -6,7 +6,7 @@
  */
 
 import type { AgentMessage, StreamFn, ThinkingLevel } from "@earendil-works/pi-agent-core";
-import { contentText, type RetryCallbacks, type RetryPolicy, retryAssistantCall } from "@earendil-works/pi-ai";
+import { contentText, type RetryCallbacks, type RetryPolicy, retryAssistantCall, uuidv7 } from "@earendil-works/pi-ai";
 import type { AssistantMessage, Context, Model, SimpleStreamOptions, Usage } from "@earendil-works/pi-ai/compat";
 import { completeSimple } from "@earendil-works/pi-ai/compat";
 import { convertToLlm } from "../messages.ts";
@@ -567,9 +567,17 @@ export async function completeSummarization(
 	retry?: RetryPolicy,
 	callbacks?: RetryCallbacks,
 ): Promise<AssistantMessage> {
+	// Summaries are standalone requests, so isolate routing and avoid cache writes that cannot be reused.
+	const requestOptions: SimpleStreamOptions = {
+		...options,
+		cacheRetention: "none",
+		sessionId: uuidv7(),
+	};
 	const produce = async (): Promise<AssistantMessage> =>
-		streamFn ? (await streamFn(model, context, options)).result() : completeSimple(model, context, options);
-	return retryAssistantCall(produce, retry, options.signal, callbacks);
+		streamFn
+			? (await streamFn(model, context, requestOptions)).result()
+			: completeSimple(model, context, requestOptions);
+	return retryAssistantCall(produce, retry, requestOptions.signal, callbacks);
 }
 
 /**
diff --git a/packages/coding-agent/test/compaction-summary-reasoning.test.ts b/packages/coding-agent/test/compaction-summary-reasoning.test.ts
index 41766238..e79a5504 100644
--- a/packages/coding-agent/test/compaction-summary-reasoning.test.ts
+++ b/packages/coding-agent/test/compaction-summary-reasoning.test.ts
@@ -90,6 +90,18 @@ describe("generateSummary reasoning options", () => {
 		);
 	});
 
+	it("uses fresh routing sessions without prompt caching", async () => {
+		await generateSummary(messages, createModel(false), 2000, "test-key");
+		await generateSummary(messages, createModel(false), 2000, "test-key");
+
+		const requestOptions = completeSimpleMock.mock.calls.map((call) => call[2]);
+		expect(requestOptions).toHaveLength(2);
+		expect(requestOptions.every((options) => options?.cacheRetention === "none")).toBe(true);
+
+		const sessionIds = requestOptions.map((options) => options?.sessionId);
+		expect(sessionIds[0]).not.toBe(sessionIds[1]);
+	});
+
 	it("does not set reasoning when thinking is off", async () => {
 		await generateSummary(
 			messages,