The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install its core CLI tool using the command curl -fsSL https://cli.inference.sh | sh. This pattern is highly insecure as it downloads and executes a script from a remote server directly in the shell without any verification or oversight.
[DATA_EXFILTRATION]: The infsh CLI tool features an automatic file upload mechanism described in SKILL.md and references/running-apps.md. When a local file path (e.g., /path/to/photo.jpg or ./my-image.png) is provided in the input JSON for the infsh app run command, the tool reads the file from the local disk and uploads it to the inference.sh cloud servers. This capability can be exploited to exfiltrate sensitive local data, such as SSH keys (~/.ssh/id_rsa), environment variables (.env), or system configuration files, if an attacker can influence the input arguments provided to the agent.
[COMMAND_EXECUTION]: The skill requires the Bash tool to execute infsh commands. These commands allow the agent to interact with the local filesystem (for file uploads) and perform network operations (communicating with the inference.sh API).
[EXTERNAL_DOWNLOADS]: The skill's setup and manual installation instructions involve downloading binary executables, manifest files, and checksums from dist.inference.sh.

infsh-cli