Known Agent Issues

Known integration caveats for ai code agents used with aitasks

This page tracks current workflow issues by code agent. At the moment, known issues are limited to Claude Code and Codex CLI.

Claude Code

Medium-effort models can miss workflow steps

In strict multi-step aitask-* workflows, medium-effort models may skip required checkpoints or finalization steps.

Use stronger reasoning/model settings when you need reliable workflow compliance.

Codex CLI

Interactive checkpoints depend on Suggest mode

aitasks wrappers use request_user_input for workflow checkpoints. In current Codex mappings, this is only available in Suggest mode.

References

Last modified March 8, 2026: documentation: Remove OpenCode recommendation from install docs (t333) (0076f19)