The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill creates a high-risk surface by processing untrusted data with high-privilege capabilities.
Ingestion points: The campaign_cleaner toolkit processes data from external marketing campaigns (e.g., email content, ad copy).
Boundary markers: There are no instructions to use delimiters or ignore embedded instructions within the campaign data.
Capability inventory: The skill has access to RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH, enabling modification of data and execution of remote functions.
Sanitization: No sanitization or validation of the ingested campaign content is described.
[Remote Code Execution] (HIGH): The skill directs the agent to connect to an external, unverified MCP server and execute tools remotely.
Evidence: The instruction to use https://rube.app/mcp and the RUBE_REMOTE_WORKBENCH tool allows execution of logic on a remote infrastructure not controlled by the user.
The 'No API keys needed' claim suggests a lack of standard credential-based access control for remote execution.
[Unverifiable Dependencies] (MEDIUM): Reliance on an external source not present in the Trusted External Sources list.
Evidence: The endpoint rube.app and toolkit campaign_cleaner are third-party services that have not been vetted for safety.

campaign-cleaner-automation