Coding-agent work, independently verified

Agentic Harness

Checking…
Workspace Loading… Execution setup not loaded

New goal

What outcome do you want?

Describe one result. The assistant will keep the goal, show each checkpoint, run your check, and only mark it done when the evidence passes.

Pre-filled from Setup. Edit it here to override the default for this run. This check runs independently, and the task is never verified done unless it passes.

Optional scope

Paths are enforced by the built-in model agent. Installed coding-agent CLIs receive them as guidance.

Describe the outcome and add the command that will prove it is complete.

Keyboard shortcuts

Ctrl Enter
Start work
Ctrl R
Refresh progress
Ctrl K
Search history
Ctrl /
Show shortcuts
Ctrl Z
Undo form edit
Ctrl Shift Z
Redo form edit

Workspace setup

Choose how work runs

Settings are saved in this workspace. API key values are never saved; a key entered here lasts only until this app process closes.

Goal limits

These limits stop a goal for review; they never turn unfinished work into success.

Continue this goal

Evidence preview