The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using npx, specifically npx skills find, npx skills add, npx skills check, and npx skills update to manage local skill installations.
[REMOTE_CODE_EXECUTION]: The skill facilitates the installation and execution of remote code via npx skills add <package>. This command downloads packages from external sources (such as GitHub) and incorporates them into the agent's environment. This is the intended primary purpose of the skill.
[EXTERNAL_DOWNLOADS]: The skill uses network-enabled commands to fetch skill definitions and packages from the https://skills.sh/ domain and various GitHub repositories, including those belonging to trusted organizations like Vercel Labs.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted user input for search queries and processes output from the npx skills find command, which could contain descriptions authored by third parties.
Ingestion points: User-provided queries in npx skills find [query] and results returned by the CLI.
Boundary markers: No explicit delimiters are defined to separate user/tool content from instructions.
Capability inventory: Extensive command execution via the npx ecosystem.
Sanitization: The instructions do not specify any sanitization or validation of the external content before it is presented to the user or used for installation.

find-skills