The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The skill documentation includes a command to download and execute a shell script from https://cli.inference.sh using the pattern curl -fsSL ... | sh. This is a highly dangerous practice as it allows an untrusted third party to execute arbitrary code with the user's local privileges without any integrity verification.
[EXTERNAL_DOWNLOADS] (HIGH): The skill promotes the installation of additional tools and scripts via npx skills add inference-sh/skills. The inference-sh organization is not on the list of trusted providers, making these external dependencies unverifiable and risky.
[COMMAND_EXECUTION] (HIGH): The skill configuration (allowed-tools) explicitly permits the execution of the Bash tool with infsh *. Combined with the unverified installation method of the infsh utility, this grants the agent the ability to run potentially malicious commands through a third-party interface.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill ingests untrusted data in the form of image_url and prompt fields. While the current capabilities are focused on image processing, there is a risk that malicious data from these inputs could influence agent behavior or be used to exploit vulnerabilities in the infsh CLI.

image-upscaling