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.
This commit is contained in:
Chinesezjc
2026-07-26 03:36:51 +08:00
parent b7322f6716
commit cb5cc7f15c
2 changed files with 10 additions and 6 deletions

View File

@@ -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 {