The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the infsh CLI tool. Command execution is appropriately restricted to the infsh binary using the allowed-tools metadata field, minimizing the potential for unauthorized system commands.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the infsh tool and related components via npx from the inference.sh platform. These are legitimate references to the service provider's infrastructure.
[PROMPT_INJECTION]: The skill contains surfaces for processing external data via model prompts.
Ingestion points: User-supplied strings are passed as the prompt argument in infsh app run commands within SKILL.md.
Boundary markers: There are no explicit delimiters or warnings to ignore instructions within the prompt strings.
Capability inventory: The skill allows execution of the infsh CLI for model inference tasks.
Sanitization: No input sanitization or escaping is performed on the user-provided prompts before execution.

nano-banana