The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to interact with package managers, including npm audit, pnpm up, and yarn install. These commands are standard for Node.js development and are used here specifically to implement security remediations within the user's project environment.
[EXTERNAL_DOWNLOADS]: The skill triggers the download and installation of packages from official registries (such as npmjs.org) via the project's existing package manager. These operations are essential for updating vulnerable dependencies to patched versions.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes untrusted data from project files to construct shell commands. This risk is inherent to the task of dependency management.
Ingestion points: The skill reads project manifests (package.json), lockfiles (pnpm-lock.yaml, yarn.lock), and the JSON output from audit commands.
Boundary markers: The instructions do not specify the use of delimiters or boundary markers when interpolating strings from project files into commands.
Capability inventory: The skill has the capability to execute a wide range of CLI commands through npm, pnpm, and yarn, including installation and script execution (package.json scripts).
Sanitization: There is no explicit requirement for the agent to sanitize package names or version strings extracted from the project environment before using them as command-line arguments.

node-dependency-vulns