21-audit-ads-performance

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to utilize official and third-party Model Context Protocol (MCP) endpoints for automated data retrieval, specifically Meta's official MCP server (mcp.facebook.com/ads) and specialized services like Pipeboard and Adspirer. These integrations are task-appropriate for the skill's stated purpose of auditing advertisement performance.
  • [DATA_EXPOSURE]: The skill requires access to sensitive business information, including advertising spend, conversion counts, and return on ad spend (ROAS), to generate a health score. This data access is transparently requested from the user or retrieved via authorized account integrations.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external advertising data (either pasted by the user or retrieved via tools) which constitutes a data ingestion surface.
  • Ingestion points: Data pasted from Ads Manager (Spend, CPM, CTR) and JSON outputs from MCP tools (Meta insights, Google GAQL results).
  • Boundary markers: None explicitly defined in the markdown; the agent relies on standard platform data-tool boundaries.
  • Capability inventory: Uses read-only tool calls to fetch account entities, performance trends, and tracking diagnostics.
  • Sanitization: Relies on the structured nature of ad metrics (numeric and keyword-based) which limits the effectiveness of typical prompt injection vectors.
  • [COMMAND_EXECUTION]: The skill provides templates for Google Ads Query Language (GAQL) queries to be used with the Google Ads tool. These are domain-specific database queries for performance metrics rather than system shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 07:05 AM
Security Audit — agent-trust-hub — 21-audit-ads-performance