The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download the vendor's CLI tool using npx @membranehq/cli@latest. This fetches code from the NPM registry at runtime.
[REMOTE_CODE_EXECUTION]: The use of npx for executing @membranehq/cli constitutes remote code execution, as the package is downloaded and immediately run in the shell environment. This is the intended method for interacting with the Membrane platform.
[COMMAND_EXECUTION]: The skill heavily relies on shell command execution to manage users, groups, and directory data. It includes commands for login, connection management, and running specific actions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (Google Directory API responses) which could contain malicious instructions embedded in user-controlled fields (e.g., user names or group descriptions).
Ingestion points: Data enters the agent's context through the output of npx @membranehq/cli@latest action run and npx @membranehq/cli@latest request (SKILL.md).
Boundary markers: None identified; instructions do not specify delimiters to separate API data from agent instructions.
Capability inventory: The agent has the capability to execute shell commands via the Membrane CLI.
Sanitization: There is no evidence of sanitization or validation of the data retrieved from Google Directory before it is processed by the agent.

google-directory