The Agent Skills Directory

Indirect Prompt Injection: The skill processes information from untrusted external sources, including web search results and uploaded financial documents. This creates a potential surface for indirect prompt injection, where instructions hidden within these sources could attempt to influence agent behavior. * Ingestion points: Sources include uploaded documents, web search results, and MCP server data as specified in Phase 1. * Boundary markers: The instructions lack explicit boundaries or warnings for the agent to disregard instructions found within the processed data. * Capability inventory: The agent can write files via the xlsx skill and perform network operations through web search. * Sanitization: No specific data sanitization or escaping procedures are defined for external content.
Dynamic Execution: The skill contains Python code templates to guide the agent in performing Excel operations. These templates are used for local script generation to facilitate the intended data pack construction.

datapack-builder