Files
codex/agentydragon/tasks/23-interactive-container-command-affordance.md
T

2.1 KiB
Raw Blame History


id: 23 title: Interactive Container Command Affordance via Hotkey status: Not started # one of: Not started, Started, Needs manual review, Done, Cancelled summary: Provide a keybinding to run arbitrary shell commands in the agent’s container and display output inline. goal: | Add a user-facing affordance (e.g. a hotkey) to invoke arbitrary shell commands within the agent's container during a session for on-demand inspection and debugging. The typed command should be captured as a chat turn, executed via the existing shell tool, and its output rendered inline in the chat UI.

Acceptance Criteria

  • Bind a hotkey (e.g. Ctrl+M) that opens a prompt for the user to type any shell command.
  • When the user submits, capture the command as if entered in the chat input, and invoke the shell tool with the command in the agent’s container.
  • Display the command invocation and its stdout/stderr output inline in the chat window, respecting formatting rules (e.g. compact rendering settings).
  • Support chaining multiple commands in separate turns; history should show these command turns normally.
  • Provide unit or integration tests simulating a user hotkey press, command input, and verifying the shell tool is called and output is displayed.

Implementation

How it was implemented

  • Define a new keybinding (configurable, default Ctrl+M) in the TUI to trigger a ShellCommandPrompt overlay.
  • In the overlay, accept arbitrary user input and dispatch it as a ToolInvocation(ShellTool, command) event in the agent’s event loop.
  • Leverage the existing shell tool backend to execute the command in the container and capture its output.
  • Render the command invocation and result inline in the chat UI using the command-rendering logic (honoring compact mode and spacing options).
  • Add integration tests to simulate the hotkey, input prompt, and verify the shell tool call and inline rendering.

Notes

  • This feature aids debugging and inspection without leaving the agent workflow.
  • Ensure that security policies (e.g. sandbox restrictions) still apply to these commands.