marketing-meta-ads-analyzer

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of Meta Ads Excel reports, which presents a surface for indirect prompt injection. \n
  • Ingestion points: User-provided .xlsx files uploaded during the workflow (SKILL.md, Step 2). \n
  • Boundary markers: Absent; the instructions do not explicitly command the agent to ignore or delimit potentially malicious instructions within the data. \n
  • Capability inventory: The agent can generate multiple file formats (HTML, DOCX, PPTX, MD) and trigger automated tool chaining to other specialized agents (SKILL.md, Step 9). \n
  • Sanitization: Absent; no specific methods for sanitizing or validating cell content are described prior to analysis. \n- [DYNAMIC_EXECUTION]: The generated HTML reports dynamically load external libraries from CDNs. \n
  • Evidence: The skill specifies the use of React and Recharts via CDN imports for rendering interactive charts in the output files (references/G.md). \n- [SAFE]: The skill demonstrates safe practices regarding external integrations and internal logic. \n
  • Official API: It supports real-time data fetching through the official Meta Ads Connector (mcp.facebook.com/ads) using OAuth 2.0 (references/H.6). \n
  • Vendor Resource Chaining: Automated chaining occurs within the vendor's ecosystem, passing data to tools such as moai-coworker and moai-writer for content refinement (SKILL.md, Step 9).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 01:42 PM
Security Audit — agent-trust-hub — marketing-meta-ads-analyzer