Skip to content

[upstream #612] agents#612: redesign stop hooks with structured exit status and per-lane state #387

Closed as not planned

Description

@berry-13

Follow-up from LibreChat-AI/agents#612: the opt-in verify-before-stop (Stop) and stop-on-passing-check (PostToolBatch) hooks were moved out of that PR because review kept finding shell forms their text heuristic misreads.

Redesign once before reopening:

  • Pass the bash tool's exit status, timeout and kill state as structured data to hooks instead of parsing output text.
  • Keep verification state per agent lane, since parallel agents share one Run and their batches overlap.
  • Count only the hook's own continuations, while honoring continuationBudgetRemaining.
  • Treat tests in the same tool batch as an edit as concurrent with it.

Open review findings to carry over: LibreChat-AI/agents#612 (comment), #discussion_r4296257768, #discussion_r4296257788, #discussion_r4296257802.

Activity

  1. berry-13 commented on Oct 16, 2026

    @berry-13
    OwnerAuthor

    Closed: no follow-up is needed for this agents change. The code this tracked was removed from LibreChat-AI/agents#612 and each review finding was answered on its thread there.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requestreadyreviewctl followups state

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions