The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses npx to download and execute code from external sources. Specifically, the add command in commands/add.md executes npx --yes skills add <owner/repo>, where the repository path is a user-controlled argument. The install command in commands/install.md also executes npx skills add vercel-labs/find-skills to install the search backend.
[EXTERNAL_DOWNLOADS]: The skill fetches executable content and metadata from the npm registry and GitHub repositories during the installation and update processes (npx skills add, npx skills check).
[COMMAND_EXECUTION]: The skill relies on the Bash tool to perform system-level operations, including Git root detection via git rev-parse --show-toplevel and package management via npx.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by processing untrusted data from skill repositories and local configuration files.
Ingestion points: The skill reads repository paths from user input and configuration from .groove/index.md and skills-lock.json.
Boundary markers: No explicit delimiters or instructions to ignore embedded instructions within processed skill data are specified.
Capability inventory: The skill possesses the ability to execute shell commands (npx, git), read and write files (skills-lock.json), and interact with the user.
Sanitization: There is no evidence of validation or sanitization of the <owner/repo> string before it is passed to the shell-executed npx command.

skills