Merge pull request #3065 from code-yeongyu/fix/issue-2984-v2

fix: reset consecutiveFailures on abort so session recovers after user cancel

src/hooks/runtime-fallback/event-handler.test.ts

@@ -104,4 +104,62 @@ describe("createEventHandler", () => {
     expect(abortCalls).toEqual([])
     expect(state.pendingFallbackModel).toBe(undefined)
   })
+
+  it("#given a cancelled session #when session.error receives an abort error #then fallback retry state is reset", async () => {
+    const sessionID = "session-cancelled"
+    const deps = createDeps()
+    const abortCalls: string[] = []
+    const clearCalls: string[] = []
+    const state = createFallbackState("google/gemini-2.5-pro")
+    state.currentModel = "openai/gpt-5.4"
+    state.fallbackIndex = 1
+    state.attemptCount = 2
+    state.pendingFallbackModel = "openai/gpt-5.4"
+    state.failedModels.set("google/gemini-2.5-pro", Date.now())
+    deps.sessionStates.set(sessionID, state)
+    deps.sessionRetryInFlight.add(sessionID)
+    deps.sessionAwaitingFallbackResult.add(sessionID)
+    deps.sessionStatusRetryKeys.set(sessionID, "retry:2")
+    const handler = createEventHandler(deps, createHelpers(deps, abortCalls, clearCalls))
+
+    await handler({ event: { type: "session.error", properties: { sessionID, error: { name: "AbortError" } } } })
+
+    const resetState = deps.sessionStates.get(sessionID)
+    expect(resetState?.originalModel).toBe("google/gemini-2.5-pro")
+    expect(resetState?.currentModel).toBe("google/gemini-2.5-pro")
+    expect(resetState?.fallbackIndex).toBe(-1)
+    expect(resetState?.attemptCount).toBe(0)
+    expect(resetState?.pendingFallbackModel).toBe(undefined)
+    expect(resetState?.failedModels.size).toBe(0)
+    expect(deps.sessionRetryInFlight.has(sessionID)).toBe(false)
+    expect(deps.sessionAwaitingFallbackResult.has(sessionID)).toBe(false)
+    expect(deps.sessionStatusRetryKeys.has(sessionID)).toBe(false)
+    expect(clearCalls).toEqual([sessionID])
+    expect(abortCalls).toEqual([])
+  })
+
+  it("#given a cancelled session #when session.idle fires #then fallback retry state stays cleared", async () => {
+    const sessionID = "session-cancelled-idle"
+    const deps = createDeps()
+    const abortCalls: string[] = []
+    const clearCalls: string[] = []
+    const state = createFallbackState("google/gemini-2.5-pro")
+    state.currentModel = "openai/gpt-5.4"
+    state.fallbackIndex = 1
+    state.attemptCount = 2
+    state.pendingFallbackModel = "openai/gpt-5.4"
+    deps.sessionStates.set(sessionID, state)
+    const handler = createEventHandler(deps, createHelpers(deps, abortCalls, clearCalls))
+
+    await handler({ event: { type: "session.error", properties: { sessionID, error: { name: "MessageAbortedError" } } } })
+    clearCalls.length = 0
+
+    await handler({ event: { type: "session.idle", properties: { sessionID } } })
+
+    const resetState = deps.sessionStates.get(sessionID)
+    expect(resetState?.currentModel).toBe("google/gemini-2.5-pro")
+    expect(resetState?.attemptCount).toBe(0)
+    expect(clearCalls).toEqual([sessionID])
+    expect(abortCalls).toEqual([])
+  })
 })

src/hooks/runtime-fallback/event-handler.ts

@@ -6,6 +6,7 @@ import { extractStatusCode, extractErrorName, classifyErrorType, isRetryableErro
 import { createFallbackState } from "./fallback-state"
 import { getFallbackModelsForSession } from "./fallback-models"
 import { SessionCategoryRegistry } from "../../shared/session-category-registry"
+import { isAbortError } from "../../shared/is-abort-error"
 import { resolveFallbackBootstrapModel } from "./fallback-bootstrap-model"
 import { dispatchFallbackRetry } from "./fallback-retry-dispatcher"
 import { createSessionStatusHandler } from "./session-status-handler"
@@ -13,6 +14,19 @@ import { createSessionStatusHandler } from "./session-status-handler"
 export function createEventHandler(deps: HookDeps, helpers: AutoRetryHelpers) {
   const { config, pluginConfig, sessionStates, sessionLastAccess, sessionRetryInFlight, sessionAwaitingFallbackResult, sessionFallbackTimeouts, sessionStatusRetryKeys } = deps
   const sessionStatusHandler = createSessionStatusHandler(deps, helpers, sessionStatusRetryKeys)
+  const cancelledSessions = new Set<string>()
+
+  const resetRetryState = (sessionID: string) => {
+    const state = sessionStates.get(sessionID)
+    if (state) {
+      sessionStates.set(sessionID, createFallbackState(state.originalModel))
+    }
+
+    sessionRetryInFlight.delete(sessionID)
+    sessionAwaitingFallbackResult.delete(sessionID)
+    sessionStatusRetryKeys.delete(sessionID)
+    helpers.clearSessionFallbackTimeout(sessionID)
+  }
 
   const handleSessionCreated = (props: Record<string, unknown> | undefined) => {
     const sessionInfo = props?.info as { id?: string; model?: string } | undefined
@@ -32,6 +46,7 @@ export function createEventHandler(deps: HookDeps, helpers: AutoRetryHelpers) {
 
     if (sessionID) {
       log(`[${HOOK_NAME}] Cleaning up session state`, { sessionID })
+      cancelledSessions.delete(sessionID)
       sessionStates.delete(sessionID)
       sessionLastAccess.delete(sessionID)
       sessionRetryInFlight.delete(sessionID)
@@ -46,28 +61,35 @@ export function createEventHandler(deps: HookDeps, helpers: AutoRetryHelpers) {
     const sessionID = props?.sessionID as string | undefined
     if (!sessionID) return
 
-    helpers.clearSessionFallbackTimeout(sessionID)
-
     if (sessionRetryInFlight.has(sessionID) || sessionAwaitingFallbackResult.has(sessionID)) {
       await helpers.abortSessionRequest(sessionID, "session.stop")
     }
 
-    sessionRetryInFlight.delete(sessionID)
-    sessionAwaitingFallbackResult.delete(sessionID)
-    sessionStatusRetryKeys.delete(sessionID)
-
-    const state = sessionStates.get(sessionID)
-    if (state?.pendingFallbackModel) {
-      state.pendingFallbackModel = undefined
-    }
+    cancelledSessions.add(sessionID)
+    resetRetryState(sessionID)
 
     log(`[${HOOK_NAME}] Cleared fallback retry state on session.stop`, { sessionID })
   }
 
+  const handleMessageUpdated = (props: Record<string, unknown> | undefined) => {
+    const info = props?.info as Record<string, unknown> | undefined
+    const sessionID = info?.sessionID as string | undefined
+    const role = info?.role as string | undefined
+    if (!sessionID || role !== "user") return
+
+    cancelledSessions.delete(sessionID)
+  }
+
   const handleSessionIdle = (props: Record<string, unknown> | undefined) => {
     const sessionID = props?.sessionID as string | undefined
     if (!sessionID) return
 
+    if (cancelledSessions.has(sessionID)) {
+      resetRetryState(sessionID)
+      log(`[${HOOK_NAME}] Cleared fallback retry state for cancelled session on idle`, { sessionID })
+      return
+    }
+
     if (sessionAwaitingFallbackResult.has(sessionID)) {
       log(`[${HOOK_NAME}] session.idle while awaiting fallback result; keeping timeout armed`, { sessionID })
       return
@@ -100,6 +122,13 @@ export function createEventHandler(deps: HookDeps, helpers: AutoRetryHelpers) {
 
     const resolvedAgent = await helpers.resolveAgentForSessionFromContext(sessionID, agent)
 
+    if (isAbortError(error)) {
+      cancelledSessions.add(sessionID)
+      resetRetryState(sessionID)
+      log(`[${HOOK_NAME}] session.error matched cancellation; cleared retry state`, { sessionID, resolvedAgent })
+      return
+    }
+
     if (sessionRetryInFlight.has(sessionID)) {
       log(`[${HOOK_NAME}] session.error skipped - retry in flight`, {
         sessionID,
@@ -176,6 +205,7 @@ export function createEventHandler(deps: HookDeps, helpers: AutoRetryHelpers) {
     if (event.type === "session.created") { handleSessionCreated(props); return }
     if (event.type === "session.deleted") { handleSessionDeleted(props); return }
     if (event.type === "session.stop") { await handleSessionStop(props); return }
+    if (event.type === "message.updated") { handleMessageUpdated(props); return }
     if (event.type === "session.idle") { handleSessionIdle(props); return }
     if (event.type === "session.status") { await sessionStatusHandler(props); return }
     if (event.type === "session.error") { await handleSessionError(props); return }

src/hooks/todo-continuation-enforcer/continuation-injection.ts

@@ -61,6 +61,11 @@ export async function injectContinuation(args: {
     return
   }
 
+  if (state?.wasCancelled) {
+    log(`[${HOOK_NAME}] Skipped injection: session was cancelled`, { sessionID })
+    return
+  }
+
   if (isContinuationStopped?.(sessionID)) {
     log(`[${HOOK_NAME}] Skipped injection: continuation stopped for session`, { sessionID })
     return
@@ -145,6 +150,11 @@ Remaining tasks:
 ${todoList}`
 
   const injectionState = sessionStateStore.getExistingState(sessionID)
+  if (injectionState?.wasCancelled) {
+    log(`[${HOOK_NAME}] Skipped injection: session was cancelled before prompt`, { sessionID })
+    return
+  }
+
   if (injectionState) {
     injectionState.inFlight = true
   }

src/hooks/todo-continuation-enforcer/handler.ts

@@ -37,7 +37,13 @@ export function createTodoContinuationHandler(args: {
       const error = props?.error as { name?: string } | undefined
       if (error?.name === "MessageAbortedError" || error?.name === "AbortError") {
         const state = sessionStateStore.getState(sessionID)
+        state.wasCancelled = true
         state.abortDetectedAt = Date.now()
+        state.lastIncompleteCount = undefined
+        state.lastInjectedAt = undefined
+        state.awaitingPostInjectionProgressCheck = false
+        state.stagnationCount = 0
+        state.consecutiveFailures = 0
         log(`[${HOOK_NAME}] Abort detected via session.error`, { sessionID, errorName: error.name })
       }
 

src/hooks/todo-continuation-enforcer/idle-event.ts

@@ -42,6 +42,11 @@ export async function handleSessionIdle(args: {
     return
   }
 
+  if (state.wasCancelled) {
+    log(`[${HOOK_NAME}] Skipped: session was cancelled`, { sessionID })
+    return
+  }
+
   if (state.abortDetectedAt) {
     const timeSinceAbort = Date.now() - state.abortDetectedAt
     if (timeSinceAbort < ABORT_WINDOW_MS) {

src/hooks/todo-continuation-enforcer/non-idle-events.ts

@@ -25,14 +25,20 @@ export function handleNonIdleEvent(args: {
           return
         }
       }
-      if (state) state.abortDetectedAt = undefined
+      if (state) {
+        state.abortDetectedAt = undefined
+        state.wasCancelled = false
+      }
       sessionStateStore.cancelCountdown(sessionID)
       return
     }
 
     if (role === "assistant") {
       const state = sessionStateStore.getExistingState(sessionID)
-      if (state) state.abortDetectedAt = undefined
+      if (state) {
+        state.abortDetectedAt = undefined
+        state.wasCancelled = false
+      }
       sessionStateStore.cancelCountdown(sessionID)
       return
     }
@@ -60,7 +66,10 @@ export function handleNonIdleEvent(args: {
     const sessionID = properties?.sessionID as string | undefined
     if (sessionID) {
       const state = sessionStateStore.getExistingState(sessionID)
-      if (state) state.abortDetectedAt = undefined
+      if (state) {
+        state.abortDetectedAt = undefined
+        state.wasCancelled = false
+      }
       sessionStateStore.cancelCountdown(sessionID)
     }
     return
@@ -70,7 +79,10 @@ export function handleNonIdleEvent(args: {
     const sessionID = properties?.sessionID as string | undefined
     if (sessionID) {
       const state = sessionStateStore.getExistingState(sessionID)
-      if (state) state.abortDetectedAt = undefined
+      if (state) {
+        state.abortDetectedAt = undefined
+        state.wasCancelled = false
+      }
       sessionStateStore.cancelCountdown(sessionID)
     }
     return

src/hooks/todo-continuation-enforcer/todo-continuation-enforcer.test.ts

@@ -1270,7 +1270,7 @@ describe("todo-continuation-enforcer", () => {
     expect(promptCalls).toHaveLength(0)
   })
 
-  test("should inject when abort flag is stale (>3s old)", async () => {
+  test("should keep skipping after cancel even when the abort window is stale", async () => {
     fakeTimers.restore()
     // given - session with incomplete todos and old abort timestamp
     const sessionID = "main-stale-abort"
@@ -1299,8 +1299,7 @@ describe("todo-continuation-enforcer", () => {
 
     await wait(3000)
 
-    // then - continuation injected (abort flag is stale)
-    expect(promptCalls.length).toBeGreaterThan(0)
+    expect(promptCalls).toHaveLength(0)
   }, { timeout: 15000 })
 
   test("should clear abort flag on user message activity", async () => {
@@ -1343,6 +1342,44 @@ describe("todo-continuation-enforcer", () => {
     expect(promptCalls.length).toBeGreaterThan(0)
   }, { timeout: 15000 })
 
+  test("should reset failure state and keep skipping after a cancelled run", async () => {
+    fakeTimers.restore()
+    const sessionID = "main-reset-after-cancel"
+    setMainSession(sessionID)
+    mockMessages = [
+      { info: { id: "msg-1", role: "user" } },
+      { info: { id: "msg-2", role: "assistant" } },
+    ]
+
+    const hook = createTodoContinuationEnforcer(createMockPluginInput(), {})
+
+    await hook.handler({
+      event: { type: "session.idle", properties: { sessionID } },
+    })
+
+    await wait(2500)
+    expect(promptCalls.length).toBeGreaterThan(0)
+
+    promptCalls.length = 0
+
+    await hook.handler({
+      event: {
+        type: "session.error",
+        properties: { sessionID, error: { name: "MessageAbortedError" } },
+      },
+    })
+
+    await wait(3100)
+
+    await hook.handler({
+      event: { type: "session.idle", properties: { sessionID } },
+    })
+
+    await wait(2500)
+
+    expect(promptCalls).toHaveLength(0)
+  }, { timeout: 15000 })
+
   test("should clear abort flag on assistant message activity", async () => {
     fakeTimers.restore()
     // given - session with abort detected
@@ -1866,4 +1903,64 @@ describe("todo-continuation-enforcer", () => {
     expect(promptCalls).toHaveLength(0)
   })
 
+  test("should reset consecutiveFailures after user-initiated abort and resume after fresh activity [regression #2984]", async () => {
+    fakeTimers.restore()
+    const sessionID = "main-abort-recovery"
+    setMainSession(sessionID)
+    const mockInput = createMockPluginInput()
+    mockInput.client.session.todo = async () => ({
+      data: [
+        { id: "1", content: "Write tests", status: "pending", priority: "high" },
+      ],
+    })
+
+    let shouldFail = true
+    let promptCallCount = 0
+    mockInput.client.session.promptAsync = async (_opts: PromptRequestOptions) => {
+      promptCallCount++
+      if (shouldFail) {
+        throw new Error("promptAsync failed (3ms) unknown error")
+      }
+      promptCalls.push({
+        sessionID: _opts.path.id,
+        agent: _opts.body.agent,
+        model: _opts.body.model,
+        text: _opts.body.parts[0].text,
+      })
+    }
+
+    const hook = createTodoContinuationEnforcer(mockInput, {})
+
+    await hook.handler({ event: { type: "session.idle", properties: { sessionID } } })
+    await wait(2500)
+    expect(promptCallCount).toBe(1)
+
+    await hook.handler({
+      event: {
+        type: "session.error",
+        properties: { sessionID, error: { name: "MessageAbortedError" } },
+      },
+    })
+
+    shouldFail = false
+    await wait(9000)
+    await hook.handler({ event: { type: "session.idle", properties: { sessionID } } })
+    await wait(2500)
+    expect(promptCallCount).toBe(1)
+
+    await hook.handler({
+      event: {
+        type: "message.updated",
+        properties: { info: { sessionID, role: "user" } },
+      },
+    })
+
+    await hook.handler({ event: { type: "session.idle", properties: { sessionID } } })
+    await wait(2500)
+
+    expect(promptCallCount).toBe(2)
+    expect(promptCalls).toHaveLength(1)
+  }, { timeout: 20000 })
+
+
 })

src/hooks/todo-continuation-enforcer/types.ts

@@ -26,6 +26,7 @@ export interface SessionState {
   countdownTimer?: ReturnType<typeof setTimeout>
   countdownInterval?: ReturnType<typeof setInterval>
   isRecovering?: boolean
+  wasCancelled?: boolean
   countdownStartedAt?: number
   abortDetectedAt?: number
   lastIncompleteCount?: number

src/hooks/unstable-agent-babysitter/index.test.ts

@@ -181,4 +181,37 @@ describe("unstable-agent-babysitter hook", () => {
     expect(promptCalls.length).toBe(1)
     Date.now = originalNow
   })
+
+  test("skips follow-up reminder after the main session is cancelled", async () => {
+    setMainSession("main-1")
+    const promptCalls: Array<{ input: unknown }> = []
+    const ctx = createMockPluginInput({
+      messagesBySession: {
+        "main-1": [
+          { info: { agent: "sisyphus", model: { providerID: "openai", modelID: "gpt-4" } } },
+        ],
+        "bg-1": [
+          { info: { role: "assistant" }, parts: [{ type: "thinking", thinking: "deep thought" }] },
+        ],
+      },
+      promptCalls,
+    })
+    const backgroundManager = createBackgroundManager([createTask()])
+    const hook = createUnstableAgentBabysitterHook(ctx, {
+      backgroundManager,
+      config: { timeout_ms: 120000 },
+    })
+    const firstNow = Date.now()
+    const originalNow = Date.now
+    let currentNow = firstNow
+    Date.now = () => currentNow
+
+    await hook.event({ event: { type: "session.idle", properties: { sessionID: "main-1" } } })
+    await hook.event({ event: { type: "session.error", properties: { sessionID: "main-1", error: { name: "AbortError" } } } })
+    currentNow += 5 * 60 * 1000 + 1
+    await hook.event({ event: { type: "session.idle", properties: { sessionID: "main-1" } } })
+
+    expect(promptCalls.length).toBe(1)
+    Date.now = originalNow
+  })
 })

src/hooks/unstable-agent-babysitter/unstable-agent-babysitter-hook.ts

@@ -2,6 +2,7 @@ import type { BackgroundManager } from "../../features/background-agent"
 import { getMainSessionID, getSessionAgent } from "../../features/claude-code-session-state"
 import { log } from "../../shared/logger"
 import { createInternalAgentTextPart, resolveInheritedPromptTools } from "../../shared"
+import { isAbortError } from "../../shared/is-abort-error"
 import {
   buildReminder,
   extractMessages,
@@ -117,17 +118,70 @@ async function getThinkingSummary(ctx: BabysitterContext, sessionID: string): Pr
 
 export function createUnstableAgentBabysitterHook(ctx: BabysitterContext, options: BabysitterOptions) {
   const reminderCooldowns = new Map<string, number>()
+  const cancelledSessions = new Set<string>()
 
   const eventHandler = async ({ event }: { event: { type: string; properties?: unknown } }) => {
+    const props = event.properties as Record<string, unknown> | undefined
+
+    if (event.type === "session.error") {
+      const sessionID = props?.sessionID as string | undefined
+      if (!sessionID || !isAbortError(props?.error)) return
+
+      cancelledSessions.add(sessionID)
+      reminderCooldowns.clear()
+      log(`[${HOOK_NAME}] Marked session cancelled`, { sessionID })
+      return
+    }
+
+    if (event.type === "session.stop") {
+      const sessionID = props?.sessionID as string | undefined
+      if (!sessionID) return
+
+      cancelledSessions.add(sessionID)
+      reminderCooldowns.clear()
+      log(`[${HOOK_NAME}] Marked session cancelled via session.stop`, { sessionID })
+      return
+    }
+
+    if (event.type === "message.updated") {
+      const info = props?.info as Record<string, unknown> | undefined
+      const sessionID = info?.sessionID as string | undefined
+      const role = info?.role as string | undefined
+      if (!sessionID || (role !== "user" && role !== "assistant")) return
+
+      cancelledSessions.delete(sessionID)
+      return
+    }
+
+    if (event.type === "tool.execute.before" || event.type === "tool.execute.after") {
+      const sessionID = props?.sessionID as string | undefined
+      if (!sessionID) return
+
+      cancelledSessions.delete(sessionID)
+      return
+    }
+
+    if (event.type === "session.deleted") {
+      const sessionInfo = props?.info as { id?: string } | undefined
+      if (!sessionInfo?.id) return
+
+      cancelledSessions.delete(sessionInfo.id)
+      return
+    }
+
     if (event.type !== "session.idle") return
 
-    const props = event.properties as Record<string, unknown> | undefined
     const sessionID = props?.sessionID as string | undefined
     if (!sessionID) return
 
     const mainSessionID = getMainSessionID()
     if (!mainSessionID || sessionID !== mainSessionID) return
 
+    if (cancelledSessions.has(mainSessionID)) {
+      log(`[${HOOK_NAME}] Skipped reminder: session was cancelled`, { sessionID: mainSessionID })
+      return
+    }
+
     const tasks = options.backgroundManager.getTasksByParentSession(mainSessionID)
     if (tasks.length === 0) return
 

src/shared/is-abort-error.ts

@@ -0,0 +1,20 @@
+export function isAbortError(error: unknown): boolean {
+  if (!error) return false
+
+  if (typeof error === "object") {
+    const errObj = error as Record<string, unknown>
+    const name = errObj.name as string | undefined
+    const message = (errObj.message as string | undefined)?.toLowerCase() ?? ""
+
+    if (name === "MessageAbortedError" || name === "AbortError") return true
+    if (name === "DOMException" && message.includes("abort")) return true
+    if (message.includes("aborted") || message.includes("cancelled") || message.includes("interrupted")) return true
+  }
+
+  if (typeof error === "string") {
+    const lower = error.toLowerCase()
+    return lower.includes("abort") || lower.includes("cancel") || lower.includes("interrupt")
+  }
+
+  return false
+}