The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, specifically Google Business Profile reviews, customer Q&A, and business updates. This establishes a surface for indirect prompt injection where malicious content embedded in these public profiles could attempt to influence the agent's behavior. However, the risk is low as the agent's capabilities are limited to data analysis and reporting.
Ingestion points: get_business_reviews, get_google_business_questions, and get_business_updates fetch content from external third parties.
Boundary markers: The skill does not define specific delimiters or instructions to distinguish untrusted data from the system prompt.
Capability inventory: The available tools are restricted to read-only SEO metrics and specific analysis functions; no filesystem writing, code execution, or non-whitelisted network capabilities are present.
Sanitization: There is no explicit sanitization or filtering logic for the ingested review or post content.
[COMMAND_EXECUTION]: The skill utilizes a set of predefined MCP tools for SEO data retrieval. These tools are task-specific and do not provide a path for arbitrary command execution or shell injection.

local-seo