The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): The skill has an indirect prompt injection surface where untrusted user input influences data processed by shell commands with high-privilege file system access.
Ingestion points: Ingestion of input_image and motion_prompt variables defined in SKILL.md.
Boundary markers: Absent; there are no delimiters or instructions to isolate user-provided content.
Capability inventory: Includes refly file download for writing to the user's desktop and the open command for process execution.
Sanitization: Absent; the FILE_NAME variable is used in shell commands without validation or escaping.
[COMMAND_EXECUTION] (MEDIUM): The skill provides automated shell scripts that interact with the local filesystem and launch applications, creating a risk if the workflow output is compromised.
[DATA_EXFILTRATION] (LOW): The skill makes network requests to refly.ai via its CLI, which is a non-whitelisted domain.

fal-video