marketing-pipeline-ai-content-automation

Warn

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to clone an external repository (github.com/pennydinh/marketing-pineline-share.git) that does not belong to the stated skill author (ara.so / Aradotso). This constitutes a dependency on an unverified third-party source.
  • [REMOTE_CODE_EXECUTION]: The installation guide requires users to execute the cloned third-party code via npm install and npm run dev. Since the source is unverified and unaffiliated with the author, this poses a risk of executing potentially malicious code within the user's environment.
  • [PROMPT_INJECTION]: The skill implements an automated research-to-content workflow that is vulnerable to indirect prompt injection. It ingests data from external, untrusted sources (Twitter, LinkedIn, and TechCrunch) and interpolates it directly into LLM prompts.
  • Ingestion points: The news-scanner.ts module fetches external articles and social media content based on keywords.
  • Boundary markers: The prompt templates in claude-generator.ts and openai-generator.ts use simple text labels (e.g., "Based on the following recent research:") which provide minimal protection against instructions embedded within the crawled content.
  • Capability inventory: The pipeline has the capability to render video content using Remotion, which involves dynamic bundling and execution of project assets via Webpack.
  • Sanitization: The provided code snippets do not implement sanitization, filtering, or validation of the content retrieved from external research sources before it is processed by the AI models.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 24, 2026, 12:52 AM
Security Audit — agent-trust-hub — marketing-pipeline-ai-content-automation