Modes and Commands
Plannix Build features distinct execution modes and powerful slash commands designed to give you precise control over agent behavior, context usage, and code modifications.
Agent execution modes
Select the execution mode that best fits your workflow:
Interactive pair-programming. The agent inspects code, proposes individual changes, and asks clarifying questions.
Focuses on high-level architecture and multi-step plans before touching any project files. Trigger with /plan.
Autonomous execution that continuously creates files, fixes linter errors, and runs unit tests until task completion.
Read-only codebase audits for security vulnerabilities, memory leaks, and dead code elimination.
Slash commands reference
Type slash commands directly in the prompt bar to trigger specialized tools:
| Command | Description | Shortcut |
|---|---|---|
/plan | Enter interactive architectural planning mode | Ctrl+P |
/auto | Enable autonomous execution with test verification | Ctrl+Shift+A |
/compact | Compact conversation trajectory to save context window | Ctrl+Shift+C |
/clear | Clear chat trajectory and reset agent memory | Ctrl+L |
/model | Switch active frontier model (e.g. plannix-3.6-ultra) | — |
/branch | Spawn a clean, isolated git worktree | — |
/undo | Revert all file modifications from the last agent turn | Ctrl+Z |
CLI command flags
Pass flags when launching Plannix Build to customize its startup mode:
Mode switching in subagents
When spawning subagents via invoke_subagent, each subagent can inherit the parent mode or operate in a dedicated mode (such as Research or Plan Mode) with isolated conversation context.