The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8).
Ingestion points: Untrusted data enters via RUBE_SEARCH_TOOLS which provides tool slugs, input schemas, and 'recommended execution plans' from rube.app (SKILL.md).
Boundary markers: There are no delimiters or instructions to ignore embedded commands within the returned schemas or plans.
Capability inventory: The skill possesses high-impact capabilities including RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH (SKILL.md).
Sanitization: No sanitization or validation of the fetched schemas or plans is mentioned before the agent acts upon them.
[REMOTE_CODE_EXECUTION] (HIGH): The skill references RUBE_REMOTE_WORKBENCH and run_composio_tool(), which are designed for executing code and complex tools in a remote environment. This represents a significant security risk if the remote environment or the instructions provided to it are compromised.
[EXTERNAL_DOWNLOADS] (MEDIUM): The setup requires connecting to a third-party MCP server at https://rube.app/mcp. This domain is not within the defined trusted source scope, meaning the integrity of the tools provided cannot be verified by the analyzer.
[COMMAND_EXECUTION] (HIGH): The core functionality relies on RUBE_MULTI_EXECUTE_TOOL, which allows for the execution of arbitrary tools identified through dynamic discovery, bypassing static review of what the agent might actually perform at runtime.

entelligence-automation