The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess calls to interact with GStreamer utilities (gst-inspect-1.0 and gst-launch-1.0). These calls are securely implemented in scripts/validate_pipeline.py and scripts/generate_pipeline.py using list-form arguments and shlex.split(), which prevents shell injection by ensuring the OS executes the command directly without a shell intermediary.
[SAFE]: Input validation is applied to pipeline strings, including a maximum length check of 16,384 characters and regex-based sanitization of element names ([a-zA-Z0-9_-]+) before they are passed to system tools.
[SAFE]: The skill relies entirely on the Python standard library for its logic, including the BM25 retrieval engine and the pipeline validator. There are no external package dependencies or remote code downloads.
[SAFE]: Pipeline generation is performed by querying a local, verified CSV dataset (data/data.csv). The retrieval logic uses BM25 scoring and structural metadata boosting to select existing, safe pipeline patterns rather than generating arbitrary code.
[SAFE]: The dry-run validation feature in scripts/validate_pipeline.py is designed to check syntax without full execution. It safely substitutes actual sources and sinks with fakesrc and fakesink and strips configuration paths to prevent unintended file or network access during the check.
[SAFE]: Administrative or persistent actions, such as writing a pipeline to a script file, are strictly user-triggered (opt-in) and use absolute paths, following the principle of least privilege and user transparency.

deepstream-generate-pipeline