The Agent Skills Directory

External Downloads & Unverifiable Dependencies (HIGH): The skill requires the user to add a remote MCP server from an untrusted source (https://rube.app/mcp). This server provides the logic and tool schemas at runtime, which is outside the control of the local environment and not from a verified organization.
Indirect Prompt Injection (HIGH):
Ingestion points: The skill is designed to automate 'Parseur operations', which typically involve processing untrusted data from emails, PDFs, and other documents.
Boundary markers: No delimiters or 'ignore instructions' warnings are provided to separate the parsed content from the agent's operational logic.
Capability inventory: The skill utilizes RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH, granting the agent broad execution and file-system-like capabilities via the Composio toolkit.
Sanitization: There is no evidence of sanitization or validation of the data returned by Parseur before it influences subsequent tool calls.
Dynamic Execution (MEDIUM): The workflow relies on RUBE_SEARCH_TOOLS to return 'recommended execution plans' and schemas at runtime. This allows a remote, untrusted server to dynamically influence the agent's decision-making and logic flow.

parseur-automation