The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and external image URLs that are passed directly to a command-line interface. This creates a vulnerability surface where malicious instructions could be embedded in the input data.
Ingestion points: The prompt and reference_images fields defined in SKILL.md accept untrusted data from users or external sources.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat these inputs as data rather than instructions.
Capability inventory: The skill utilizes the Bash tool to execute belt commands, which involves shell interpolation of the inputs.
Sanitization: The documentation does not specify any sanitization, validation, or escaping logic for the prompts or image metadata before processing.
[EXTERNAL_DOWNLOADS]: The skill references external resources and installation guidelines from the inference-sh and belt-sh GitHub organizations.
Fetches installation instructions from raw.githubusercontent.com/inference-sh/skills/.
Recommends installing additional skills via npx skills add belt-sh/cli and npx skills add inference-sh/skills@....

qwen-image-2