The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands for environment setup, authentication, and service interaction.
Evidence: Commands include npm install -g @membranehq/cli@latest, membrane login, and membrane action run.
[EXTERNAL_DOWNLOADS]: The skill requires downloading a CLI tool from the public NPM registry.
Evidence: npm install -g @membranehq/cli@latest and npx @membranehq/cli@latest.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes data from external HERE APIs and uses natural language intents to discover or generate actions.
Ingestion points: Data returned from HERE services via membrane action run and descriptions used in membrane action list.
Boundary markers: None explicitly defined in the instructions to separate external data from agent instructions.
Capability inventory: The skill has the capability to install software via npm and execute dynamic actions via the membrane CLI.
Sanitization: No explicit sanitization or validation steps are documented for the data received from external sources.

here