Core Workflows
Deep Planning
/deep-planning turns DevcoreAI into a meticulous architect — investigation first, code second.
When to Use Deep Planning
Standard Plan mode is great for medium-sized tasks. Deep planning is designed for situations where jumping straight to code would be a mistake:
The 4 Phases
1. Silent Investigation
DevcoreAI explores your codebase structure and patterns without asking questions — reading files, understanding dependencies, mapping affected components.
2. Discussion
After investigation, the agent asks targeted clarifying questions about requirements, constraints, and edge cases it couldn't resolve from code alone.
3. Plan Creation
Generates implementation_plan.md with: feature overview, file-by-file changes, dependencies between changes, edge cases, and a testing strategy.
4. Task Creation
Creates a new task with trackable implementation steps derived from the plan. You can review and edit the plan before the agent starts coding.
How to Invoke
Type /deep-planning followed by a description of what you want to build. You can be in either Plan or Act mode — deep planning works in both.
Examples
/deep-planning Add multi-tenancy support to the database layer
/deep-planning Migrate the frontend from Redux to Zustand
/deep-planning Add real-time collaboration using WebSockets
The implementation_plan.md File
The output is a structured Markdown file saved in your workspace root. It contains:
- •Feature overview and scope definition
- •File-by-file changes with descriptions of what to add, modify, or remove
- •Dependencies between changes (what must happen first)
- •Edge cases and error handling considerations
- •Testing strategy for the implementation
You can edit implementation_plan.md before the agent starts coding. Changes you make to the plan file are respected during implementation.