The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill accepts user-supplied text via the --prompt argument and external images via the --start-frame and --end-frame arguments (which support URLs) in scripts/generation_api.py.
Boundary markers: The instructions in SKILL.md lack explicit boundary markers or instructions for the agent to ignore potentially embedded commands within the prompt or external image metadata.
Capability inventory: The skill possesses network capabilities through the requests library in scripts/generation_api.py and is intended to be run via shell execution.
Sanitization: Data from external sources is interpolated directly into API payloads without sanitization or filtering for instructional patterns.
[DATA_EXFILTRATION]: The skill performs network requests to an external domain https://giggle.pro. While this is the legitimate endpoint for the service provided by the skill author, it involves sending user-provided prompts and potentially sensitive image data to an external server.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands (python3 scripts/generation_api.py) with arguments derived from user input, which is a standard but inherently privileged operation.

giggle-generation-video