CLI Command Reference

Relevant source files

The following files were used as context for generating this wiki page:

• src/core/agent.ts
• src/commands/help.ts
• src/core/slashCommands.ts
• src/commands/README.md
• prd/command-ux-spec.md
• src/commands/skills-new.ts
• src/commands/automode.ts

The Autohand CLI Command Reference provides a comprehensive guide to the interactive and non-interactive commands used to manage the AI coding agent. The system is built on a modular architecture where commands are categorized into three distinct layers: slash commands for system control, at-file references for context injection, and shell commands for direct environment interaction.

Sources: prd/command-ux-spec.md:5-10, src/commands/README.md:1-5

Command Architecture and Processing

The CLI utilizes a centralized commandCompletion orchestrator to track the active input mode based on user prefixes. When a user enters a prefix, the corresponding processor handles the execution logic.

Input Processing Flow

The following diagram illustrates how the system determines which processor should handle a specific input string.

The input capture phase delegates context to specific hooks like useAtCompletion for path suggestions or slashCommandProcessor for system actions.
Sources: prd/command-ux-spec.md:46-55, src/core/agent.ts:474-485

Core Components

Component	Responsibility	Source
slashCommandProcessor	Loads built-in command descriptors and metadata.	prd/command-ux-spec.md:14-23
atCommandProcessor	Resolves relative paths, respects .gitignore, and injects file snippets.	prd/command-ux-spec.md:25-31
shellCommandProcessor	Executes commands using runCommand and streams output to Ink UI.	prd/command-ux-spec.md:33-38
SlashCommandHandler	Manages stateful command execution (e.g., model selection, memory).	src/core/agent.ts:187-205

Slash Command Reference

Slash commands are system-level instructions used to configure the agent, manage sessions, and control skills. These commands are defined in src/core/slashCommands.ts and implemented as individual modules.

Sources: src/core/slashCommands.ts:7-41, src/commands/README.md:11-23

General System Commands

Command	Alias	Description	Implementation File
/help	/?	Describes available commands and provides usage tips.	src/commands/help.ts
/model	-	Opens the configuration menu for LLM providers and models.	src/commands/model.js
/quit	/exit	Saves the session and terminates the CLI application.	src/commands/quit.js
/status	-	Displays the current workspace status and token usage.	src/commands/status.js
/undo	-	Reverts the last file mutation performed by the agent.	src/commands/undo.js

Sources: src/commands/help.ts:13-25, src/core/slashCommands.ts:32-70, src/core/agent.ts:466-470

Skill and Agent Management

Commands for extending agent capabilities and managing sub-agents.

• /skills: Lists and manages installed skills.
• /skills new: Interactive wizard to create a new skill from a description. It generates a SKILL.md file with YAML frontmatter.
• /agents: Lists available specialized sub-agents.
• /automode: Controls autonomous development loops, allowing commands like status, pause, and resume.

Sources: src/commands/skills-new.ts:21-26, src/commands/automode.ts:16-20, src/core/slashCommands.ts:65-68

Session and Memory Commands

• /sessions: Lists all past saved sessions.
• /resume : Reloads a previous conversation state by its unique session ID.
• /memory: View or manage facts and preferences stored in project/user-level memory.
• /export: Exports the current session history to Markdown, JSON, or HTML.

Sources: src/commands/README.md:19-21, src/core/agent.ts:413-417, src/core/slashCommands.ts:55-63

Advanced Command Features

Autonomous Loops (Auto-Mode)

Auto-mode enables iterative improvement cycles where the agent works autonomously toward a defined prompt.

The /automode command supports subcommands like pause to halt the loop and status to view files modified and current iteration counts.
Sources: src/commands/automode.ts:40-75, src/commands/automode.ts:137-145

Skill Generation Process

The /skills new command initiates an LLM-driven flow to create reusable instruction packages.

1. Input: User provides a name and a natural language description.
2. Selection: User chooses storage level (project or user).
3. Validation: System checks for name validity and similarity to existing skills.
4. Generation: LLM generates a SKILL.md with structured YAML frontmatter.
5. Confirmation: User previews the content before saving to the .autohand/skills directory.

Sources: src/commands/skills-new.ts:39-65, src/commands/skills-new.ts:121-135

Summary

The CLI command system provides a robust interface for both manual control and autonomous operation. By partitioning functionality into slash commands, at-references, and shell execution, it allows developers to maintain flow while providing the agent with precise context and instructions. The integration of session resumption and skill management ensures that project-specific knowledge persists across multiple interactions.

Sources: src/core/agent.ts:601-615, prd/command-ux-spec.md:57-60