The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages the ATLASCLOUD_API_KEY environment variable, which is transmitted to the official api.atlascloud.ai endpoints for authentication. This is standard and expected behavior for the skill's purpose.
[INDIRECT_PROMPT_INJECTION]: The generate command accepts a JSON file for model parameters. While this data is included in an API request, the tool performs validation against a live schema and requires an explicit --confirm-paid flag for execution.
Ingestion points: Parameters are read from a file or stdin in scripts/atlas_media.py through the _load_params function.
Boundary markers: None for the content of the prompt strings.
Capability inventory: Network requests (POST/GET) are performed using the standard library's urllib.request in scripts/atlas_media.py.
Sanitization: The tool implements type, range, and enum validation in the _validate_params method based on the retrieved model schema.
[DYNAMIC_EXECUTION]: The script dynamically fetches a JSON schema from a remote URL to determine validation rules and API paths (submit_path, result_path). These paths are safely combined with the base API URL to ensure requests are directed to the correct service domain.

atlas-cloud-media