Ling GUI Agent

Use scripts/run.py to run the project bundled in assets/ling-gui-agent/. Keep the runner in the foreground so framework cancellation reaches the complete child process group.

Configuration

Use the skill-root .env by default. For an externally managed secret file, use --config /private/path/ling.env; LING_CONFIG_FILE is also supported. Precedence is --config, LING_CONFIG_FILE, then the skill-root .env.

The selected file must define LING_BASE_URL, LING_MODEL, and LING_API_KEY. LING_BASE_URL must be the real base URL of an OpenAI-compatible API and normally has the form https://<host>/v1. Empty, malformed, or placeholder values such as https://example.com/v1 are not usable. If the value is invalid, do not guess or substitute an endpoint; ask the user for the exact OpenAI-compatible API base URL supplied by their service provider.

Never print configuration contents or pass credentials as command-line arguments. If the file or another required value is missing, do not start the GUI agent. Tell the user to copy .env.example to a private file, fill in the required values, and retry.

Run

From the skill root, run a bounded foreground task:

python3 scripts/run.py "<goal>" --platform desktop --max-steps 100