The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends installing its CLI tool via curl -fsSL https://cli.inference.sh | sh. This pattern is highly risky as it executes arbitrary code from the internet without prior inspection, which could lead to a full system compromise if the source is compromised.- [EXTERNAL_DOWNLOADS]: The installation process downloads a binary from dist.inference.sh. While the skill mentions SHA-256 verification, this remains an external dependency managed outside the skill's environment and the official package registries.- [COMMAND_EXECUTION]: The skill makes extensive use of the infsh command to run remote applications and process images. This tool must be pre-installed and authorized, and its usage involves executing shell commands based on the skill's instructions.- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It takes user-provided text (descriptions, shot details) into prompts sent to an AI image generation model (falai/flux-dev-lora) without sanitization or boundary markers.
Ingestion points: Prompt parameters within the infsh app run commands in SKILL.md.
Boundary markers: None present to distinguish instructions from user data.
Capability inventory: Execution of infsh CLI tool for remote application runs and image processing.
Sanitization: No escaping or validation is performed on the user-provided prompt strings.

storyboard-creation