The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends installing the CLI tool via curl -fsSL https://cli.inference.sh | sh. This pattern involves piping remote content directly into a shell interpreter, which allows for the execution of unverified code from an external server.
[DATA_EXFILTRATION]: The infsh CLI automatically uploads local files when a file path is provided in an app's input parameters. This creates a data exfiltration surface where sensitive local data (such as configuration files or credentials) could be uploaded to the vendor's infrastructure if the agent is directed to process those paths.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to run the infsh CLI. Although restricted to the infsh command via the allowed-tools configuration, the tool's capabilities (running AI apps, uploading files, and network communication) grant the agent significant autonomy.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external inputs for AI apps. Ingestion points: App input fields (SKILL.md, running-apps.md). Boundary markers: Absent. Capability inventory: Bash tool with file-read and network-send capabilities. Sanitization: Absent.

infsh-cli