The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the infsh CLI tool to manage user sessions and run language model applications on the Inference.sh platform.
[EXTERNAL_DOWNLOADS]: The documentation recommends the installation of related skills using npx skills add inference-sh/skills, which refers to packages within the vendor's own ecosystem.
[PROMPT_INJECTION]: The skill accepts untrusted user input to be used as prompts for external language models, which is an inherent surface for indirect prompt injection. Ingestion points: Untrusted data is passed to the LLM via the --input argument in command examples. Boundary markers: The provided examples do not use delimiters or instructions to prevent the LLM from obeying instructions embedded in the user data. Capability inventory: The skill has access to the Bash(infsh *) tool to execute remote operations. Sanitization: There is no evidence of input validation or sanitization within the skill's defined instructions.

llm-models