[BUG] Claude Code produces insufficient plans for complex multi-task projects

Here's the issue in the requested format:

Environment

• Platform (select one):
• [x] Anthropic API
• [ ] AWS Bedrock
• [ ] Google Vertex AI
• [ ] Other: <!-- specify -->
• Claude CLI version: 1.0.43 (Claude Code)
• Operating System: mac OS (latest)
• Terminal: zsh

Bug Description

Claude Code produces overly concise, implementation-focused plans when asked to create detailed architectural documents for complex projects. The same Opus model generates comprehensive plans in other interfaces (e.g., Cursor), but Claude Code's output lacks the necessary detail for breaking down complex features into manageable tasks.

Steps to Reproduce

1. Use Claude Code to request a detailed architectural plan for a complex system (e.g., "Create a comprehensive architectural plan for a multi-tenant SaaS application with real-time features")
2. Ask the same question to Opus model in Cursor or Claude web interface
3. Compare the depth, structure, and usefulness of the planning outputs

Expected Behavior

Claude Code should generate comprehensive planning documents that:

• Include detailed system architecture, component breakdowns, and implementation phases
• Are thorough enough to serve as reference for task decomposition
• Match the quality of plans produced by the same model in other interfaces

Actual Behavior

• Plans are extremely concise and lack architectural detail
• Output is implementation-focused rather than design-focused
• Plans are insufficient for breaking down complex projects into discrete tasks
• Quality is significantly lower than the same model's output in other interfaces

Additional Context

Why this matters for my workflow:

• I work on complex features that require detailed planning documents
• I break down plans into discrete tasks to avoid context pollution
• Without proper plans, I cannot effectively use Claude Code for the full development lifecycle
• Currently forced to use other tools for planning, then switch to Claude Code for implementation

This appears to be a system prompt optimization issue where Claude Code is tuned for direct code generation rather than comprehensive planning, even when explicitly requested.