[OPUS 4.5 on Claude Max] - no amount of proper planning and documentation is helping Claude complete the plan, including proper prompting, skills and other support provided to complete, verify and confirm.

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

I asked to implement a well documented set of instructions. Asked Claude to plan first using a skill I built, reviewed and approved the plan. Claude created todos, I asked Claude to refine the todos to ensure success, it did. Yet, look at the screenshot.

observability-n-resilience-system-prompt.md

<img width="683" height="115" alt="Image" src="https://github.com/user-attachments/assets/f679600f-c230-4754-aa7f-61e4b1db358a" />

What Claude Actually Did

See the image. Essentially half baked solution. This continues to happen with every task, it never finishes a plan it started, with everything provided upfront.

<img width="683" height="115" alt="Image" src="https://github.com/user-attachments/assets/f5e12647-8e96-433b-990f-e0fd6af93480" />

Expected Behavior

follow the plan and complete it fully.

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

create a nice prompt, even take Claude to do it. use a custom skill or available skill to develop a plan basis prompt, refine the plan, refine Claude code's todos and ask it to execute. it still provides a half-done solution.

<img width="683" height="115" alt="Image" src="https://github.com/user-attachments/assets/0551d6de-fd76-4852-b9ea-ecb636f7c9bc" />

Claude Model

Opus

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.0.75

Platform

Anthropic API

Additional Context

_No response_