fix(coding-agent): avoid retrying successful overflow compaction

closes #5720

packages/coding-agent/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Fixed successful overflow-triggered auto-compaction to avoid retrying completed assistant responses ([#5720](https://github.com/earendil-works/pi/issues/5720)).
+
 ## [0.79.7] - 2026-06-18
 
 ### New Features

packages/coding-agent/src/core/agent-session.ts

@@ -1824,8 +1824,17 @@ export class AgentSession {
 			return false;
 		}
 
-		// Case 1: Overflow - LLM returned context overflow error
+		// Case 1: Overflow - LLM returned context overflow error, or reported usage exceeded
+		// the configured window. A successful response over the configured window should compact
+		// but must not retry: the assistant answer already completed and agent.continue() cannot
+		// continue from an assistant message.
 		if (sameModel && isContextOverflow(assistantMessage, contextWindow)) {
+			const willRetry = assistantMessage.stopReason !== "stop";
+
+			if (!willRetry) {
+				return await this._runAutoCompaction("overflow", false);
+			}
+
 			if (this._overflowRecoveryAttempted) {
 				this._emit({
 					type: "compaction_end",
@@ -1846,7 +1855,7 @@ export class AgentSession {
 			if (messages.length > 0 && messages[messages.length - 1].role === "assistant") {
 				this.agent.state.messages = messages.slice(0, -1);
 			}
-			return await this._runAutoCompaction("overflow", true);
+			return await this._runAutoCompaction("overflow", willRetry);
 		}
 
 		// Case 2: Threshold - context is getting large

packages/coding-agent/test/suite/agent-session-compaction.test.ts

@@ -248,6 +248,37 @@ describe("AgentSession compaction characterization", () => {
 		);
 	});
 
+	it("compacts successful overflow responses without retrying", async () => {
+		const harness = await createHarness({
+			settings: { compaction: { enabled: true, keepRecentTokens: 1, reserveTokens: 0 } },
+			models: [{ id: "faux-1", contextWindow: 1, maxTokens: 100 }],
+			extensionFactories: [
+				(pi) => {
+					pi.on("session_before_compact", async (event) => ({
+						compaction: {
+							summary: "successful overflow compacted",
+							firstKeptEntryId: event.preparation.firstKeptEntryId,
+							tokensBefore: event.preparation.tokensBefore,
+							details: {},
+						},
+					}));
+				},
+			],
+		});
+		harnesses.push(harness);
+		harness.setResponses([fauxAssistantMessage("completed answer")]);
+
+		await expect(harness.session.prompt("hello")).resolves.toBeUndefined();
+
+		const compactionEnd = harness.eventsOfType("compaction_end").at(-1);
+		expect(compactionEnd).toMatchObject({
+			reason: "overflow",
+			aborted: false,
+			willRetry: false,
+		});
+		expect(harness.faux.state.callCount).toBe(1);
+	});
+
 	it("ignores stale pre-compaction assistant usage on pre-prompt checks", async () => {
 		const harness = await createHarness();
 		harnesses.push(harness);