Repository navigation
Keep the agent working after it advances the task - #435
Merged
Merged
Conversation
On a mid-turn move (the user says `/advance`), the agent's latest phase guidance was the per-turn briefing for the *old* phase — for a user-advanced phase, "stop and hand back to the user" — and the operation skill said the move "starts a new turn". Agents obeyed: they announced the new phase and stopped until prompted. - The MCP `apply_operation`/`set_state` tools now return the entered phase's briefing (`Workflow.entry_briefing`), led by an instruction to begin that phase's work in the same turn. - The rendered operation skills (claude + codex) drop "starts a new turn" for "follow the returned briefing and keep going". - The user-advanced stop clause adds: when the user asks you to advance, do it and carry straight on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tildesrc
marked this pull request as ready for review
October 1, 2026 20:03
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Agents would advance a task (e.g. the user says
/advanceout of PLANNING), announce "The task is now in ITERATING … next I'll start implementing", and stop until prompted again.Why: on a mid-turn move, the latest phase guidance in the agent's context was the per-turn briefing the UserPromptSubmit hook printed for the old phase. For a user-advanced phase that ends in "stop and hand back to the user … Don't advance on your own." The
apply_operationtool returned only the raw task, and the operation skill said the move "starts a new turn". Every signal said to stop.Changes
apply_operation/set_state(MCP) now return the task plus abriefingfor the phase just entered (Workflow.entry_briefing→TaskService.entry_briefing), led by "Begin this phase's work now, in this same turn." Terminal states get the plain terminal briefing. The REST endpoints are unchanged.Plan: the task's
plan.mdartifact.🤖 Generated with Claude Code