The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The skill is primarily a wrapper for installing and running remote code via the npx skills add command. The explicit instruction to use the -y flag allows the agent to execute arbitrary external packages without any human-in-the-loop confirmation.
[EXTERNAL_DOWNLOADS] (HIGH): The skill encourages downloading code from untrusted external sources, specifically allowing any GitHub repository (<owner/repo@skill>) to be used as a source for executable skills.
[COMMAND_EXECUTION] (HIGH): The skill provides instructions for the agent to execute shell commands (npx, bash) to perform package management tasks. When combined with unverified remote sources, this provides a direct path for system compromise.
[PROMPT_INJECTION] (HIGH): This skill exposes a massive Indirect Prompt Injection surface (Category 8). It ingests untrusted user input as search queries, searches a public ecosystem for matching code, and then provides the mechanism to execute that code. Ingestion Point: npx skills find [query]. Boundary Markers: None. Capability Inventory: Command execution (npx), remote download, and installation. Sanitization: None. A malicious actor could name a skill to match common queries and gain execution on the user's system.

find-skills