The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @membranehq/cli package from the official NPM registry. This is the vendor's command-line tool necessary for the skill's functionality and is considered a trusted resource.
[COMMAND_EXECUTION]: The instructions involve executing various subcommands of the membrane CLI to authenticate, manage connections, and run actions. These are legitimate operations performed through the platform's official tooling.
[PROMPT_INJECTION]: The skill interacts with external data from Breezy HR (e.g., candidate notes and position details), which presents a surface for indirect prompt injection.
Ingestion points: Data returned from membrane action run commands, such as candidate profiles and activity streams.
Boundary markers: Not explicitly mentioned in the skill instructions.
Capability inventory: The skill allows the agent to execute platform commands and create new actions via the Membrane CLI.
Sanitization: No specific sanitization or validation logic is defined within the instructions, relying on standard agent guardrails.

breezy