This page is intended for AI Coding Agents, not people.

Guardrails

Use these guardrails for every generated task script.

Always-safe behavior

• Prefer receive/monitor actions over transmit actions unless TX is explicitly required.
• Keep each task bounded by timeout, stop condition, or explicit cancel path.
• Use defensive parsing for every CAT response.
• Fail gracefully with clear logs instead of throwing uncaught errors.

TX/RX safety

• If script issues TX;, it must include an RX; cleanup path.
• Place RX cleanup in finally when possible.
• If TX duration is prompt-driven, enforce a hard maximum duration.
• On parse error or timeout during TX flow, default to RX cleanup.

Timer safety

• Keep references to every timeout/interval handle.
• Clear all active timers on completion and cancel.
• Avoid nested interval creation that can multiply over time.
• Prefer recursive timeout over interval when pacing depends on device responsiveness.

Retry and timeout policy

• Use bounded retries with explicit max attempts.
• Use finite wait durations for CAT operations.
• Log both retry count and terminal failure reason.
• Stop retries early on explicit user cancel.

Cancellation policy

• Support cancellation for long-running operations.
• Route cancellation through one cleanup function.
• Cleanup function should be idempotent (safe to call multiple times).

User input blocking

• Prefer async execution of long running processes.
• Use pause API to allow user to cancel script.

GUI feedback

• print API can display two lines of up to 4 characters.
• prefer setDescription API for longer messages.
• prefer setTitle API to append additional value to task title, with setTitle( context.task.title + ${newValue} )