The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill instructs the user/agent to run npx @wp-playground/cli@latest. This command fetches and executes code directly from the npm registry at runtime. While WordPress Playground is a legitimate project, the use of npx with an unpinned version from an external repository carries inherent risks of supply chain compromise or runtime code injection.
[REMOTE_CODE_EXECUTION] (MEDIUM): The run-blueprint command supports loading JSON configuration files from remote URLs (e.g., https://example.com/blueprint.json). These blueprints can define steps to install WordPress plugins from arbitrary ZIP file URLs (e.g., pluginZipFile: { "resource": "url", "url": "..." }), effectively allowing a remote file to control code installation on the local environment.
[COMMAND_EXECUTION] (MEDIUM): The skill provides instructions for executing shell commands that mount local directories (--mount) into a WebAssembly-based WordPress environment. This could lead to sensitive local data exposure if malicious paths are provided to the mount command.
[PROMPT_INJECTION] (LOW): This skill is vulnerable to Indirect Prompt Injection (Category 8) due to the ingestion of external data. \n
Ingestion points: Remote JSON blueprints and external plugin ZIP files specified in blueprints. \n
Boundary markers: Absent; there are no instructions to the agent to treat blueprint content as data only. \n
Capability inventory: Command execution (npx), local filesystem access (--mount), and network requests for site setup. \n
Sanitization: Absent; the skill does not mention validation or sandboxing of the blueprint sources or plugin URLs.

wp-playground