fix(delegate-task): remove kimi from unstable agent detection, harden callID resolution for metadata store

- Remove kimi from auto-detected unstable agents in category-resolver (only gemini/minimax remain)
- Add resolveCallID() to safely resolve callID/callId/call_id variants from tool context
- Use resolveCallID across all 5 delegate task execution paths (sync, background, unstable, continuations)
- Update writing category test to verify kimi runs sync when kimi provider is available
- Add atlas metadata preservation tests confirming tool-execute-after does not clobber metadata

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
This commit is contained in:
YeonGyu-Kim
2026-04-05 15:30:36 +09:00
parent ccfc54ab38
commit d5dfaaa3ad
10 changed files with 196 additions and 42 deletions
+91
View File
@@ -248,6 +248,91 @@ describe("atlas hook", () => {
cleanupMessageStorage(sessionID) cleanupMessageStorage(sessionID)
}) })
test("should preserve metadata when transforming output for boulder orchestrator", async () => {
// given - Atlas caller with boulder state and metadata containing sessionId
const sessionID = "session-metadata-preserve-test"
setupMessageStorage(sessionID, "atlas")
const planPath = join(TEST_DIR, "metadata-plan.md")
writeFileSync(planPath, "# Plan\n- [ ] Task 1")
const state: BoulderState = {
active_plan: planPath,
started_at: "2026-01-02T10:00:00Z",
session_ids: ["session-1"],
plan_name: "metadata-plan",
}
writeBoulderState(TEST_DIR, state)
const hook = createAtlasHook(createMockPluginInput())
const output = {
title: "Sisyphus Task",
output: `Task completed
<task_metadata>
session_id: ses_subagent_abc
</task_metadata>`,
metadata: {
sessionId: "ses_subagent_abc",
agent: "sisyphus-junior",
category: "quick",
truncated: false,
} as Record<string, unknown>,
}
// when
await hook["tool.execute.after"](
{ tool: "task", sessionID },
output
)
// then - output is transformed but metadata is preserved
expect(output.output).toContain("SUBAGENT WORK COMPLETED")
expect(output.metadata.sessionId).toBe("ses_subagent_abc")
expect(output.metadata.agent).toBe("sisyphus-junior")
expect(output.metadata.category).toBe("quick")
expect(output.metadata.truncated).toBe(false)
cleanupMessageStorage(sessionID)
})
test("should preserve metadata when appending standalone verification reminder", async () => {
// given - Atlas caller without boulder state, metadata containing sessionId
const sessionID = "session-standalone-metadata-test"
setupMessageStorage(sessionID, "atlas")
const hook = createAtlasHook(createMockPluginInput())
const output = {
title: "Sisyphus Task",
output: `Task completed
<task_metadata>
session_id: ses_standalone_def
</task_metadata>`,
metadata: {
sessionId: "ses_standalone_def",
agent: "sisyphus-junior",
model: { providerID: "openai", modelID: "gpt-5.4" },
truncated: false,
} as Record<string, unknown>,
}
// when
await hook["tool.execute.after"](
{ tool: "task", sessionID },
output
)
// then - standalone verification appended but metadata preserved
expect(output.output).toContain("LYING")
expect(output.metadata.sessionId).toBe("ses_standalone_def")
expect(output.metadata.agent).toBe("sisyphus-junior")
expect(output.metadata.model).toEqual({ providerID: "openai", modelID: "gpt-5.4" })
expect(output.metadata.truncated).toBe(false)
cleanupMessageStorage(sessionID)
})
test("should still transform when plan is complete (shows progress)", async () => { test("should still transform when plan is complete (shows progress)", async () => {
// given - boulder state with complete plan, Atlas caller // given - boulder state with complete plan, Atlas caller
const sessionID = "session-complete-plan-test" const sessionID = "session-complete-plan-test"
@@ -2103,10 +2188,14 @@ session_id: ses_untrusted_999
let nextFakeId = 99000 let nextFakeId = 99000
const originalSetTimeout = globalThis.setTimeout const originalSetTimeout = globalThis.setTimeout
const originalClearTimeout = globalThis.clearTimeout const originalClearTimeout = globalThis.clearTimeout
const originalDateNow = Date.now
let fakeNow = 0
beforeEach(() => { beforeEach(() => {
capturedTimers.clear() capturedTimers.clear()
nextFakeId = 99000 nextFakeId = 99000
fakeNow = 10000
Date.now = () => fakeNow
globalThis.setTimeout = ((callback: Function, delay?: number, ...args: unknown[]) => { globalThis.setTimeout = ((callback: Function, delay?: number, ...args: unknown[]) => {
const normalized = typeof delay === "number" ? delay : 0 const normalized = typeof delay === "number" ? delay : 0
@@ -2131,12 +2220,14 @@ session_id: ses_untrusted_999
afterEach(() => { afterEach(() => {
globalThis.setTimeout = originalSetTimeout globalThis.setTimeout = originalSetTimeout
globalThis.clearTimeout = originalClearTimeout globalThis.clearTimeout = originalClearTimeout
Date.now = originalDateNow
}) })
async function firePendingTimers(): Promise<void> { async function firePendingTimers(): Promise<void> {
for (const [id, entry] of capturedTimers) { for (const [id, entry] of capturedTimers) {
if (!entry.cleared) { if (!entry.cleared) {
capturedTimers.delete(id) capturedTimers.delete(id)
fakeNow += 6000
await entry.callback() await entry.callback()
} }
} }
@@ -3,6 +3,7 @@ import type { ExecutorContext, ParentContext } from "./executor-types"
import { storeToolMetadata } from "../../features/tool-metadata-store" import { storeToolMetadata } from "../../features/tool-metadata-store"
import { formatDetailedError } from "./error-formatting" import { formatDetailedError } from "./error-formatting"
import { getSessionTools } from "../../shared/session-tools-store" import { getSessionTools } from "../../shared/session-tools-store"
import { resolveCallID } from "./resolve-call-id"
export async function executeBackgroundContinuation( export async function executeBackgroundContinuation(
args: DelegateTaskArgs, args: DelegateTaskArgs,
@@ -37,8 +38,9 @@ export async function executeBackgroundContinuation(
}, },
} }
await ctx.metadata?.(bgContMeta) await ctx.metadata?.(bgContMeta)
if (ctx.callID) { const callID = resolveCallID(ctx)
storeToolMetadata(ctx.sessionID, ctx.callID, bgContMeta) if (callID) {
storeToolMetadata(ctx.sessionID, callID, bgContMeta)
} }
return `Background task continued. return `Background task continued.
+4 -2
View File
@@ -4,6 +4,7 @@ import type { FallbackEntry } from "../../shared/model-requirements"
import { getTimingConfig } from "./timing" import { getTimingConfig } from "./timing"
import { buildTaskPrompt } from "./prompt-builder" import { buildTaskPrompt } from "./prompt-builder"
import { storeToolMetadata } from "../../features/tool-metadata-store" import { storeToolMetadata } from "../../features/tool-metadata-store"
import { resolveCallID } from "./resolve-call-id"
import { formatDetailedError } from "./error-formatting" import { formatDetailedError } from "./error-formatting"
import { getSessionTools } from "../../shared/session-tools-store" import { getSessionTools } from "../../shared/session-tools-store"
import { SessionCategoryRegistry } from "../../shared/session-category-registry" import { SessionCategoryRegistry } from "../../shared/session-category-registry"
@@ -132,8 +133,9 @@ export async function executeBackgroundTask(
metadata, metadata,
} }
await ctx.metadata?.(unstableMeta) await ctx.metadata?.(unstableMeta)
if (ctx.callID) { const callID = resolveCallID(ctx)
storeToolMetadata(ctx.sessionID, ctx.callID, unstableMeta) if (callID) {
storeToolMetadata(ctx.sessionID, callID, unstableMeta)
} }
const taskMetadataBlock = sessionId const taskMetadataBlock = sessionId
+1 -1
View File
@@ -234,7 +234,7 @@ Available categories: ${categoryNames.join(", ")}`,
} }
const resolvedModel = actualModel?.toLowerCase() const resolvedModel = actualModel?.toLowerCase()
const isUnstableAgent = resolved.config.is_unstable_agent ?? (resolvedModel ? resolvedModel.includes("gemini") || resolvedModel.includes("minimax") || resolvedModel.includes("kimi") : false) const isUnstableAgent = resolved.config.is_unstable_agent ?? (resolvedModel ? resolvedModel.includes("gemini") || resolvedModel.includes("minimax") : false)
const defaultProviderID = categoryModel?.providerID const defaultProviderID = categoryModel?.providerID
?? parseModelString(actualModel ?? "")?.providerID ?? parseModelString(actualModel ?? "")?.providerID
@@ -0,0 +1,40 @@
import { describe, test, expect } from "bun:test"
import { resolveCallID } from "./resolve-call-id"
import type { ToolContextWithMetadata } from "./types"
describe("resolveCallID", () => {
function makeCtx(overrides: Partial<ToolContextWithMetadata> = {}): ToolContextWithMetadata {
return {
sessionID: "ses_test",
messageID: "msg_test",
agent: "sisyphus",
abort: new AbortController().signal,
...overrides,
}
}
test("#given callID is set #then returns callID", () => {
const ctx = makeCtx({ callID: "call_abc" })
expect(resolveCallID(ctx)).toBe("call_abc")
})
test("#given only callId is set #then returns callId", () => {
const ctx = makeCtx({ callId: "call_def" })
expect(resolveCallID(ctx)).toBe("call_def")
})
test("#given only call_id is set #then returns call_id", () => {
const ctx = makeCtx({ call_id: "call_ghi" })
expect(resolveCallID(ctx)).toBe("call_ghi")
})
test("#given callID and callId are both set #then prefers callID", () => {
const ctx = makeCtx({ callID: "preferred", callId: "fallback" })
expect(resolveCallID(ctx)).toBe("preferred")
})
test("#given no call ID variants are set #then returns undefined", () => {
const ctx = makeCtx()
expect(resolveCallID(ctx)).toBeUndefined()
})
})
@@ -0,0 +1,5 @@
import type { ToolContextWithMetadata } from "./types"
export function resolveCallID(ctx: ToolContextWithMetadata): string | undefined {
return ctx.callID ?? ctx.callId ?? ctx.call_id
}
+4 -2
View File
@@ -2,6 +2,7 @@ import type { DelegateTaskArgs, ToolContextWithMetadata } from "./types"
import type { ExecutorContext, SessionMessage } from "./executor-types" import type { ExecutorContext, SessionMessage } from "./executor-types"
import { isPlanFamily } from "./constants" import { isPlanFamily } from "./constants"
import { storeToolMetadata } from "../../features/tool-metadata-store" import { storeToolMetadata } from "../../features/tool-metadata-store"
import { resolveCallID } from "./resolve-call-id"
import { getTaskToastManager } from "../../features/task-toast-manager" import { getTaskToastManager } from "../../features/task-toast-manager"
import { getAgentToolRestrictions } from "../../shared/agent-tool-restrictions" import { getAgentToolRestrictions } from "../../shared/agent-tool-restrictions"
import { getMessageDir } from "../../shared" import { getMessageDir } from "../../shared"
@@ -78,8 +79,9 @@ export async function executeSyncContinuation(
}, },
} }
await ctx.metadata?.(syncContMeta) await ctx.metadata?.(syncContMeta)
if (ctx.callID) { const callID = resolveCallID(ctx)
storeToolMetadata(ctx.sessionID, ctx.callID, syncContMeta) if (callID) {
storeToolMetadata(ctx.sessionID, callID, syncContMeta)
} }
const allowTask = isPlanFamily(resumeAgent) const allowTask = isPlanFamily(resumeAgent)
+4 -2
View File
@@ -3,6 +3,7 @@ import type { DelegateTaskArgs, ToolContextWithMetadata, DelegatedModelConfig }
import type { ExecutorContext, ParentContext } from "./executor-types" import type { ExecutorContext, ParentContext } from "./executor-types"
import { getTaskToastManager } from "../../features/task-toast-manager" import { getTaskToastManager } from "../../features/task-toast-manager"
import { storeToolMetadata } from "../../features/tool-metadata-store" import { storeToolMetadata } from "../../features/tool-metadata-store"
import { resolveCallID } from "./resolve-call-id"
import { subagentSessions, syncSubagentSessions, setSessionAgent } from "../../features/claude-code-session-state" import { subagentSessions, syncSubagentSessions, setSessionAgent } from "../../features/claude-code-session-state"
import { log } from "../../shared/logger" import { log } from "../../shared/logger"
import { SessionCategoryRegistry } from "../../shared/session-category-registry" import { SessionCategoryRegistry } from "../../shared/session-category-registry"
@@ -114,8 +115,9 @@ export async function executeSyncTask(
}, },
} }
await ctx.metadata?.(syncTaskMeta) await ctx.metadata?.(syncTaskMeta)
if (ctx.callID) { const callID = resolveCallID(ctx)
storeToolMetadata(ctx.sessionID, ctx.callID, syncTaskMeta) if (callID) {
storeToolMetadata(ctx.sessionID, callID, syncTaskMeta)
} }
const promptError = await deps.sendSyncPrompt(client, { const promptError = await deps.sendSyncPrompt(client, {
+33 -25
View File
@@ -2368,41 +2368,49 @@ describe("sisyphus-task", () => {
expect(result).toContain("Artistry result here") expect(result).toContain("Artistry result here")
}, { timeout: 20000 }) }, { timeout: 20000 })
test("writing category (kimi) with run_in_background=false should force background but wait for result", async () => { test("writing category (kimi) with run_in_background=false should run sync when kimi provider is available", async () => {
// given - writing uses kimi-for-coding/k2p5 // given - writing uses kimi model which is no longer considered unstable
// Override provider cache to include kimi-for-coding provider
providerModelsSpy.mockReturnValue({
models: {
anthropic: ["claude-opus-4-6", "claude-sonnet-4-6", "claude-haiku-4-5"],
google: ["gemini-3.1-pro", "gemini-3-flash"],
openai: ["gpt-5.4", "gpt-5.3-codex"],
"kimi-for-coding": ["k2p5"],
},
connected: ["anthropic", "google", "openai", "kimi-for-coding"],
updatedAt: "2026-01-01T00:00:00.000Z",
})
cacheSpy.mockReturnValue(["anthropic", "google", "openai", "kimi-for-coding"])
const { createDelegateTask } = require("./tools") const { createDelegateTask } = require("./tools")
let launchCalled = false let launchCalled = false
let promptCalled = false
const launchedTask = {
id: "task-writing",
sessionID: "ses_writing_gemini",
description: "Writing gemini task",
agent: "sisyphus-junior",
status: "running",
}
const mockManager = { const mockManager = {
launch: async () => { launch: async () => {
launchCalled = true launchCalled = true
return launchedTask return { id: "should-not-be-called", sessionID: "x", description: "x", agent: "x", status: "running" }
}, },
getTask: () => launchedTask,
} }
const promptMock = async () => {
promptCalled = true
return { data: {} }
}
const mockClient = { const mockClient = {
app: { agents: async () => ({ data: [] }) }, app: { agents: async () => ({ data: [] }) },
config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) }, config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
model: { list: async () => [{ provider: "google", id: "gemini-3-flash" }] },
session: { session: {
get: async () => ({ data: { directory: "/project" } }), get: async () => ({ data: { directory: "/project" } }),
create: async () => ({ data: { id: "ses_writing_gemini" } }), create: async () => ({ data: { id: "ses_writing_kimi" } }),
prompt: async () => ({ data: {} }), prompt: promptMock,
promptAsync: async () => ({ data: {} }), promptAsync: promptMock,
messages: async () => ({ messages: async () => ({
data: [ data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "Writing result here" }] }]
{ info: { role: "assistant", time: { created: Date.now() } }, parts: [{ type: "text", text: "Writing result here" }] }
]
}), }),
status: async () => ({ data: { "ses_writing_gemini": { type: "idle" } } }), status: async () => ({ data: { "ses_writing_kimi": { type: "idle" } } }),
}, },
} }
@@ -2418,10 +2426,10 @@ describe("sisyphus-task", () => {
abort: new AbortController().signal, abort: new AbortController().signal,
} }
// when - writing category (gemini-3-flash) // when - writing category (kimi) with run_in_background=false
const result = await tool.execute( const result = await tool.execute(
{ {
description: "Test writing forced background", description: "Test writing sync",
prompt: "Write something", prompt: "Write something",
category: "writing", category: "writing",
run_in_background: false, run_in_background: false,
@@ -2430,10 +2438,10 @@ describe("sisyphus-task", () => {
toolContext toolContext
) )
// then - should launch as background BUT wait for and return actual result // then - should run sync, NOT forced to background (kimi is not unstable)
expect(launchCalled).toBe(true) expect(launchCalled).toBe(false)
expect(result).toContain("SUPERVISED TASK COMPLETED") expect(promptCalled).toBe(true)
expect(result).toContain("Writing result here") expect(result).not.toContain("SUPERVISED TASK COMPLETED")
}, { timeout: 20000 }) }, { timeout: 20000 })
test("is_unstable_agent=true should force background but wait for result", async () => { test("is_unstable_agent=true should force background but wait for result", async () => {
@@ -4,6 +4,7 @@ import { DEFAULT_SYNC_POLL_TIMEOUT_MS, getTimingConfig } from "./timing"
import { buildTaskPrompt } from "./prompt-builder" import { buildTaskPrompt } from "./prompt-builder"
import { cancelUnstableAgentTask } from "./cancel-unstable-agent-task" import { cancelUnstableAgentTask } from "./cancel-unstable-agent-task"
import { storeToolMetadata } from "../../features/tool-metadata-store" import { storeToolMetadata } from "../../features/tool-metadata-store"
import { resolveCallID } from "./resolve-call-id"
import { formatDuration } from "./time-formatter" import { formatDuration } from "./time-formatter"
import { formatDetailedError } from "./error-formatting" import { formatDetailedError } from "./error-formatting"
import { getSessionTools } from "../../shared/session-tools-store" import { getSessionTools } from "../../shared/session-tools-store"
@@ -81,8 +82,9 @@ export async function executeUnstableAgentTask(
}, },
} }
await ctx.metadata?.(bgTaskMeta) await ctx.metadata?.(bgTaskMeta)
if (ctx.callID) { const callID = resolveCallID(ctx)
storeToolMetadata(ctx.sessionID, ctx.callID, bgTaskMeta) if (callID) {
storeToolMetadata(ctx.sessionID, callID, bgTaskMeta)
} }
const startTime = new Date() const startTime = new Date()