launch-meta-ad-campaign
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled JavaScript utility (
tools/meta_marketing_api.js) to interact with the Meta Marketing API. These commands are used locally to prepare and publish campaign configurations and do not involve remote code execution. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, including the Meta Marketing API and content from user-provided destination URLs.
- Ingestion points: Ad account metadata from Meta and HTML content fetched from destination URLs during the preflight check stage.
- Boundary markers: The skill uses a fixed 'findings template' (Stage 3) to present external data to the user, providing a clear structure that separates external content from agent instructions.
- Capability inventory: Execution of local Node.js scripts, writing JSON state files to the local file system, and making network requests to Meta's API and user-defined URLs.
- Sanitization: The bundled tool includes rigorous Server-Side Request Forgery (SSRF) protections, specifically the
isPrivateIpandassertPublicHttpsUrlfunctions, which prevent the agent from being coerced into accessing internal or private network resources. - [SAFE]: The skill follows secure credential management practices by retrieving sensitive access tokens from environment variables (
META_MARKETING_ACCESS_TOKEN) rather than requesting them in prompts or hardcoding them in files.
Audit Metadata