The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands using the 'npx skills' CLI tool to manage agent extensions.
Evidence: Operational steps include running npx skills find, npx skills check, and npx skills update within the host environment.
[EXTERNAL_DOWNLOADS]: The skill downloads software packages from the npm registry and arbitrary GitHub repositories.
Evidence: The command npx skills add <owner/repo@skill> is used to fetch code from remote sources, and the skill references the external browse domain skills.sh.
[REMOTE_CODE_EXECUTION]: Facilitates the installation and potential execution of remote code from unverified third-party repositories.
Evidence: The skill provides instructions to run npx skills add with the -g (global) and -y (yes) flags, which installs software globally and bypasses confirmation prompts, creating a direct vector for unverified code execution.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via manipulated metadata in the search results from external registries.
Ingestion points: Results from the npx skills find command (containing third-party skill names and descriptions) are processed and displayed by the agent.
Boundary markers: Absent; there are no instructions to wrap or isolate search result text before it is presented to the user or context.
Capability inventory: The agent has shell execution capabilities via the npx command.
Sanitization: Absent; the agent is instructed to present skill information directly from the registry output without validation or filtering.

find-skills