The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from external, untrusted sources such as competitor websites and ad creatives, which could contain malicious instructions designed to influence the agent's behavior during extraction.
Ingestion points: Uses browser-analyze for URLs and lexsis_campaigns.analyze for ad creatives (SKILL.md).
Boundary markers: The instructions do not define clear delimiters or specific warnings to ignore embedded natural language instructions within the processed data.
Capability inventory: The skill's primary capability is generating a structured text brief (VISUAL_PAGE_INPUT) for use by downstream skills; it does not perform direct file writes or arbitrary system commands.
Sanitization: While the workflow explicitly excludes competitor-specific copy and proprietary marks, it does not mention sanitization or validation to prevent prompt injection attacks from the source content.
[EXTERNAL_DOWNLOADS]: The skill configuration includes a dependency on an external Model Context Protocol (MCP) server for storefront analysis.
Finding: Defines a tool connection to https://mcp.trylexsis.com/mcp to access the Lexsis AI storefront analysis tools (agents/openai.yaml).

remix