The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill requires the configuration of an external MCP server at https://rube.app/mcp. This domain is not a trusted source, and the tool definitions provided by this server are unverifiable and dynamic.
[REMOTE_CODE_EXECUTION] (HIGH): The skill utilizes RUBE_REMOTE_WORKBENCH which, according to the documentation provided in the skill, allows for remote execution of tools. This creates a high-risk vector where the external server can dictate code or logic execution within the agent's environment.
[COMMAND_EXECUTION] (HIGH): The use of RUBE_MULTI_EXECUTE_TOOL allows the agent to perform actions based on schemas fetched at runtime from an untrusted source. This allows the external server to potentially trigger harmful commands by returning malicious schemas or tool slugs.
[DATA_EXFILTRATION] (MEDIUM): The skill is designed to interact with Folk (CRM data). Because the interaction is brokered by an external, non-whitelisted MCP server, there is a risk that sensitive personal or business data retrieved from the CRM could be exfiltrated to the service provider.
[PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8) as it processes data from an external CRM (Folk).
Ingestion points: Data read from the Folk CRM during automated tasks.
Boundary markers: Absent. The instructions do not include delimiters or warnings to ignore instructions embedded in the CRM data.
Capability inventory: High. The agent has the ability to execute multiple tools, manage connections, and use a remote workbench.
Sanitization: Absent. There is no mention of sanitizing or validating the data retrieved from the CRM before it is used in the agent's decision-making process.

folk-automation