marketing-content-pipeline-automation

Pass

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated content research functionality.
  • Ingestion points: The researchTopic function in src/lib/research/crawler.ts crawls content from external, untrusted sources such as TechCrunch, Twitter, and LinkedIn.
  • Boundary markers: The instructions and code snippets lack specific delimiters or system instructions to the AI agent to ignore instructions embedded within the crawled research data.
  • Capability inventory: The skill environment has capabilities to make external network requests (Facebook Graph API), perform file system operations (video rendering), and execute subprocesses (via Remotion's reliance on ffmpeg).
  • Sanitization: There is no evidence of input validation, escaping, or filtering of the externally sourced content before it is interpolated into the prompts for Claude or OpenAI.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone a project repository from a public GitHub repository.
  • [COMMAND_EXECUTION]: The documentation provides instructions for installing dependencies using package managers (npm, yarn, pnpm) and running a local development server.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 29, 2026, 07:10 AM
Security Audit — agent-trust-hub — marketing-content-pipeline-automation