The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of an external package and references installation guidelines from a remote repository.- Evidence: Instructions to run npx skills add belt-sh/cli to install the required CLI tool.- Evidence: Reference to https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md for installation instructions.- [PROMPT_INJECTION]: The skill accepts arbitrary text prompts and images from users to generate or edit content, which serves as a potential surface for indirect prompt injection.- Ingestion points: The prompt and reference_images input parameters used in command examples and defined in the options table in SKILL.md.- Boundary markers: Inputs are encapsulated within JSON strings passed to the belt CLI tool, but the skill lacks explicit 'ignore embedded instructions' markers.- Capability inventory: The skill uses Bash(belt *) to execute shell commands and interact with a remote inference API.- Sanitization: There is no evidence of input validation, escaping, or filtering of the external user-provided content before it is processed.

qwen-image-2