The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a reference for the astria command-line tool, which is executed via Bash to interact with the Astria API for image and video generation, model management, and workspace administration.
[DATA_EXFILTRATION]: The skill specifies that the astria tool manages credentials via ~/.astria/config.json, which is the standard configuration directory for this vendor's CLI. It also provides instructions for handing off sessions using astria agent handoff, which can upload files and skills to the vendor's platform. The instructions include explicit safety warnings against including credentials, hidden reasoning, or agent configuration directories in these handoffs.
[INDIRECT_PROMPT_INJECTION]: The skill describes operations that ingest untrusted data, such as user-provided prompt text and external image/video URLs. These inputs are processed by the astria tool to perform network operations (generation/upload) and file system writes (downloads).
Ingestion points: Prompts provided via --text and --video-prompt arguments; external media via --image-url, --input-video, and --reference parameters in various subcommands.
Capability inventory: The astria tool performs network requests to the Astria API, reads local files for upload (tunes create, video, variate), and writes generated assets to the local filesystem (download).
Boundary markers: No explicit prompt delimiters or "ignore" instructions are specified for user-provided text interpolation.
Sanitization: No explicit sanitization or validation of the prompt content is mentioned within the reference instructions.

astria-api