The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the @membranehq/cli package from the NPM registry. This is an official tool provided by the vendor (Membrane) to facilitate platform interactions.
[COMMAND_EXECUTION]: The skill relies on executing membrane CLI commands to perform operations such as authentication, connection management, and API requests. These commands are executed in the local environment to interface with the Membrane service.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it retrieves data from an external CRM (EspoCRM). Content within leads, tasks, or opportunities could potentially contain malicious instructions intended to influence the agent's behavior.
Ingestion points: Data retrieved via membrane action run commands (e.g., list-leads, get-task) as described in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are provided in the skill documentation.
Capability inventory: The agent has the ability to execute shell commands (membrane) and perform network requests as detailed in SKILL.md.
Sanitization: No sanitization or validation of the retrieved CRM data is described.
[SAFE]: The skill follows security best practices by using a CLI-based authentication flow, ensuring that API keys and sensitive credentials are not stored or handled directly within the skill's code or prompts.

espocrm