[MODEL] Ignores user instructions, makes assumptions without verification

Type of Behavior Issue

Claude ignored my instructions or configuration

What You Asked Claude to Do

Follow the project's CLAUDE.md rule: verify with actual execution before declaring anything complete.

What Claude Actually Did

Repeatedly made assumptions and proceeded without verification, despite the rule being loaded into context.

Expected Behavior

Verify each step before declaring "done" or "safe", as instructed.

Impact

Critical - Pattern observed across multiple sessions over at least the past week. This session alone: 1+ hour wasted, erroneous data uploaded to my cloud storage. Cumulative impact across sessions is significantly larger.

Claude Code Version

2.1.96

Platform

Anthropic API

Claude Model

Opus