From cb5cc7f15c9dde80c93517e2bb38fb68b1e45f5e Mon Sep 17 00:00:00 2001
From: Chinesezjc <jczhai@deepseek.com>
Date: Sun, 26 Jul 2026 03:36:51 +0800
Subject: [PATCH] test(todo): align real-model e2e prompt and scenario with
 parallel todos

The key-gated todo_write e2e still pinned the previous contract: its
TODO_SYSTEM_PROMPT instructed at most one in_progress task and the
scenario never exercised parallel active tasks. Update the prompt to the
new guidance and make the scenario record two simultaneously in_progress
tasks; verified against the real API.
---
 examples/headless-agent/tests/harness.ts        |  5 +++--
 examples/headless-agent/tests/todo-write.e2e.ts | 11 +++++++----
 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/examples/headless-agent/tests/harness.ts b/examples/headless-agent/tests/harness.ts
index e1edc1dadd..45eb3f8ee2 100644
--- a/examples/headless-agent/tests/harness.ts
+++ b/examples/headless-agent/tests/harness.ts
@@ -28,8 +28,9 @@ export const SYSTEM_PROMPT = 'You are a coding agent. Use bash for file operatio
 /** System prompt for the todo_write e2e: nudges the model to plan with the tool. */
 export const TODO_SYSTEM_PROMPT = 'You are a coding agent. For multi-step work, '
   + 'use the todo_write tool to track a task list: send the WHOLE list each call, '
-  + 'keep at most one task in_progress (exactly one while work remains), and mark '
-  + 'a task completed as soon as it is done.'
+  + 'mark every task being actively worked on in_progress (several at once when '
+  + 'work runs in parallel, at least one while work remains), and mark a task '
+  + 'completed as soon as it is done.'
 
 /** Options for {@link codingHarness}. */
 export interface CodingHarnessOptions {
diff --git a/examples/headless-agent/tests/todo-write.e2e.ts b/examples/headless-agent/tests/todo-write.e2e.ts
index c3053572d9..83571b1305 100644
--- a/examples/headless-agent/tests/todo-write.e2e.ts
+++ b/examples/headless-agent/tests/todo-write.e2e.ts
@@ -29,9 +29,10 @@ describe.skipIf(!process.env.DEEPSEEK_API_KEY)('todo_write: real model records a
     const agent = ctx.agentLoop.create(SessionId('e2e-todo'), { provider: 'deepseek', model: 'deepseek-v4-flash' })
 
     agent.followup([{ type: 'text', text:
-      'Use the todo_write tool to record a plan of exactly two steps: first '
-      + '"inspect the failing test" (in_progress), then "apply the fix" (pending). '
-      + 'Send both in one todo_write call, then reply with the single word DONE.' }])
+      'Use the todo_write tool to record a plan of exactly three steps for work '
+      + 'running in parallel: "inspect the failing test" (in_progress), '
+      + '"watch the background build" (in_progress), then "apply the fix" (pending). '
+      + 'Send all three in one todo_write call, then reply with the single word DONE.' }])
     await waitForIdle(ctx, agent)
 
     const events = [...agent.session.events]
@@ -40,13 +41,15 @@ describe.skipIf(!process.env.DEEPSEEK_API_KEY)('todo_write: real model records a
     const calls = events.filter(event => event.type === 'tool/call')
     expect(calls.some(event => event.data.name === 'todo_write')).toBe(true)
 
-    // And the tool wrote a todo/write event to the log — verify the WORLD.
+    // And the tool wrote a todo/write event to the log — verify the WORLD,
+    // including two simultaneously in_progress tasks (the parallel contract).
     const todoEvents = events.filter(event => event.type === 'todo/write')
     expect(todoEvents.length).toBeGreaterThan(0)
 
     const todos = (todoEvents.at(-1)!).data.todos
     expect(todos).toEqual([
       { content: 'inspect the failing test', status: 'in_progress' },
+      { content: 'watch the background build', status: 'in_progress' },
       { content: 'apply the fix', status: 'pending' },
     ])
   }, 120_000)