Skip to content
Refact Documentation

Shell Tool

The Shell Tool allows the Refact.ai Agent to execute any command-line tool with user confirmation directly from the chat interface. This integration is ideal for automating tasks while ensuring safety and user control.

Basic Configurations

Timeout

  • Specifies the maximum time (in seconds) that a command is allowed to run
  • If the command exceeds this time, it is automatically terminated
  • Any output (stdout/stderr) is collected and presented to the user

Confirmation Rules

Provides a safety mechanism for executing potentially destructive or sensitive commands:

Ask User

  • Commands matching patterns in this list will prompt the user for confirmation before execution
  • By default, the wildcard (*) matches all commands

Deny

  • Commands matching patterns in this list will be automatically blocked
  • Example: sudo* blocks commands requiring elevated privileges

Users can add or remove rows to customize these rules according to their preferences and security requirements.

Advanced Configuration Options

Output Filter

Controls how the output of executed commands is processed and displayed:

Basic Limits

  • Limit Lines: Restricts the output to a specified number of lines (e.g., 100)
  • Limit Characters: Restricts the output to a maximum number of characters (e.g., 10,000)

Output Processing

  • Valuable Top or Bottom: Determines whether the tool prioritizes the start (top) or end (bottom) of the output for relevance
  • Grep: Uses a regular expression (e.g., (?i)error) to filter and highlight specific content in the output
  • Grep Context Lines: Defines the number of surrounding lines to include with matches from the grep filter
  • Remove from Output: Allows for removing unwanted patterns or content from the displayed output

These settings help manage large or verbose outputs, focusing only on the most critical information.