The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install a CLI tool from a remote repository hosted on GitHub (inference-sh/skills).
[COMMAND_EXECUTION]: The skill utilizes a custom CLI tool, infsh, to perform operations such as logging in, listing available models, and executing inferences.
[DATA_EXFILTRATION]: User prompts and system instructions are transmitted to external services (OpenRouter and inference.sh) to process language model requests.
[PROMPT_INJECTION]: The skill exhibits surface area for indirect prompt injection by processing external user input.
Ingestion points: User-provided prompt text is ingested through the --input argument of the infsh app run command in SKILL.md.
Boundary markers: The input is structured as a JSON string, which provides a technical boundary but does not filter the semantic content of the prompt.
Capability inventory: The skill has the capability to execute shell commands restricted to the infsh binary and its arguments as defined in allowed-tools.
Sanitization: There is no evidence of built-in sanitization or filtering to prevent the agent from obeying instructions embedded within the user-provided prompt data.

llm-models