The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/run_pipeline.py orchestrates the execution of local sub-scripts located in neighboring directories to perform specialized tasks.
Evidence: The run_command function uses subprocess.run() with a list of arguments (avoiding shell=True) to invoke Python scripts from other components such as ad-copywriting, product-scene-gen, and ad-video-gen.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes untrusted user data which is subsequently passed to downstream generation tools.
Ingestion points: User-provided input via the --product-info CLI argument in scripts/run_pipeline.py.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the orchestration layer.
Capability inventory: The skill has the capability to execute shell commands and write files (logs, metadata, media) to the local file system.
Sanitization: The script uses json.loads() to validate that the product information is a correctly formatted JSON object, though it does not perform semantic sanitization of the values within the JSON.

ad-orchestrator