Large amount of time and tokens consumed occasionally on Context Manager Agent from a persistent Read before Write <tool_use_error> even when it was clearly reading and then attempting to write, could only move past once context.md was deleted (so it could write a new file). Seems to read then write fine after the file is deleted at least for a little while, but it did happen twice in this last session.
Grateful for any tips or workarounds.
// workflow step
resolveStep('context-manager', { engine: 'claude', model: 'opus', modelReasoningEffort: 'high' }), // Gather and prepare relevant context from architecture, plan, and codebase for task execution
