The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill has a high-risk attack surface because it instructs the agent to ingest and follow "recommended execution plans" and tool schemas from an external source (RUBE_SEARCH_TOOLS).
Ingestion points: Output from the RUBE_SEARCH_TOOLS call as described in SKILL.md.
Boundary markers: Absent. The instructions tell the agent to "Always search tools first" and use the results without verification.
Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH provide the capability to execute tool sequences and code in a remote environment.
Sanitization: Absent. The skill directs the agent to use exact field names and types from search results.
[External Downloads] (MEDIUM): The skill requires the configuration of an external third-party MCP server (https://rube.app/mcp). This endpoint dynamically defines the tools and behaviors available to the agent at runtime.
[Remote Code Execution] (MEDIUM): The use of RUBE_REMOTE_WORKBENCH indicates capabilities for running remote tasks or scripts, which increases the impact if the agent follows a malicious execution plan.
[Command Execution] (MEDIUM): RUBE_MULTI_EXECUTE_TOOL allows for the automated execution of various tools with parameters fetched dynamically from the network.

seismic-automation