[MODEL] Model fails to follow explicit sequential instructions in CLAUDE.md after recent update

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude ignored my instructions or configuration

What You Asked Claude to Do

User prompt (verbatim):

"Hi Claude. Please read CLAUDE.md and follow the directions precisely."

CLAUDE.md instructions (verbatim):

Before starting any development work or research: Read PRODUCT_VISION.md
Then: Go to docs/development/README.md

What Claude Actually Did

What the model does instead:

Skips PRODUCT_VISION.md entirely and jumps directly to README.md and/or CURRENT_TASK.md. The model demonstrably read and processed the instruction, then ignored it.

Expected Behavior

Expected behavior: Read PRODUCT_VISION.md first, then README.md, then follow the hierarchy.

Files Affected

**Attempted to read**: CURRENT_TASK.md *First*
**Read**: README.md

**Skipped**: PRODUCT_VISION.md

Permission Mode

Accept Edits was OFF (manual approval required)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

_No response_

Claude Model

Sonnet

Relevant Conversation

Prompt: 


P05-S02-WU2F-T05

Hi Claude. Please read CLAUDE.md and **follow the directions precisely**.


Response:


I'll read the project documentation to get oriented.

Read CURRENT_TASK.md
Read README.md
CURRENT_TASK.md doesn't exist yet. Let me check CURRENT_WORK.md and the Task Workflow.

Impact

High - Significant unwanted changes

Claude Code Version

2.1.73

Platform

Anthropic API

Additional Context

This is recent. It is not only this, but it also has started completely ignoring stop gates within my workflow documents. VERY FRUSTRATING!!!!