The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill defines allowed-tools: Bash(npx skills *). This wildcard permission allows the agent to execute any subcommand of the 'skills' CLI, including those that modify the system or install new software, without further restricted scoping.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill provides instructions for npx skills add, which downloads and installs external packages from the inference.sh registry. As this source is not on the trusted organizations list, these dependencies are considered unverifiable.
[REMOTE_CODE_EXECUTION] (MEDIUM): By design, the skill's purpose is to fetch and install new capabilities (code) at runtime. This creates a remote code execution vector where the security of the agent becomes dependent on the integrity of the third-party registry and the specific packages requested.

related-skill