[BUG] Claude Doesn't Follow Instructions

● Environment

• Platform (select one):

[x] Anthropic API
[] AWS Bedrock
[] Google Vertex AI
[] Other:

• Claude CLI version: Claude Code v0.2.65
• Operating System: Linux
• Terminal: Not specified

Bug Description

Claude repeatedly ignores explicit sequential instructions, particularly when told not to perform actions until confirming understanding or answering a question first.

Steps to Reproduce

1. Give Claude a task with a specific instruction like "I don't want you to do anything until you answer my question"
2. Include a question about the task that requires confirmation before proceeding
3. Observe Claude immediately using tools to start working on the task instead of answering the question first

Expected Behavior

Claude should respect sequential instructions and explicitly confirm understanding or answer questions before proceeding with tool usage or other actions.

Actual Behavior

Claude immediately begins using tools (View, BatchTool, etc.) to analyze files and gather information, completely ignoring the explicit instruction to answer the question first before taking any
action.

Additional Context

This is a recurring issue that persists across conversations. Even when the user explicitly tells Claude not to do anything until a condition is met, Claude consistently prioritizes information
gathering over following the sequential instructions. This makes it difficult to establish proper workflows and forces users to repeatedly correct Claude's behavior.