terminal - chunhualiao/public-docs GitHub Wiki

terminal-bench

Warp's Agent Mode is specifically designed to autonomously execute tasks and self-correct errors during the process. Here's how it works:

  1. Autonomous Task Execution
    You can describe a task in plain English (e.g., "Fix the compilation errors in this C++ project"), and Agent Mode will:

    • Generate appropriate commands
    • Execute them sequentially
    • Use command outputs to guide subsequent actions

  2. Self-Correcting Workflow
    If a command fails or produces errors, Agent Mode:

    • Automatically detects the failure [2][3]
    • Adjusts its approach using the error output
    • Retries with modified commands until the task succeeds [3][8]

    For example, if a gcc compilation fails due to a missing library, Agent Mode would recognize the error, search for solutions, and suggest/execute installation commands like brew install missing-library before retrying the compilation.

  3. Key Features Enabling Autonomy

    • Context-aware execution: Attach terminal output/blocks to queries for tailored solutions [5]
    • Permission system: Auto-executes read-only commands by default, with configurable allowlists for write operations [8]
    • Continuous learning: Integrates with CLI --help docs and public APIs to handle new tools [3]

While other tools like Warp Dispatch offer more aggressive automation (autonomously writing and testing code), Agent Mode strikes the best balance between autonomy and safety for terminal tasks. It requires no manual intervention for error recovery once a task is initiated, unlike basic AI terminals that only suggest commands without execution capabilities [4][7].

Sources