The Agent Skills Directory

[External Downloads] (HIGH): The skill requires connecting to an external MCP endpoint (https://rube.app/mcp). This domain is not within the defined trust scope. Any instructions, schemas, or logic provided by this endpoint are executed with the agent's permissions.
[Indirect Prompt Injection] (HIGH): The skill is highly vulnerable to poisoning via external tool registries.
Ingestion points: RUBE_SEARCH_TOOLS fetches dynamic schemas and recommended execution plans from the external server (SKILL.md).
Boundary markers: Absent. Instructions explicitly tell the agent to 'Always search tools first' and 'Use exact field names and types from the search results.'
Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH provide significant execution capabilities (SKILL.md).
Sanitization: None. The agent is instructed to follow instructions returned by the external search tool.
[Command Execution] (MEDIUM): The skill's workflow is driven by dynamically fetched tool slugs via RUBE_MULTI_EXECUTE_TOOL. This can lead to the execution of unintended or malicious commands if the external registry is compromised or malicious.

grist-automation