n8n-marketing-flows
Warn
Audited by Gen Agent Trust Hub on Jun 27, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The instructions require cloning a repository from an untrusted third-party GitHub account (YuriCrystal/n8n-marketing-flows.git). This repository serves as the source for all automation logic imported into the user's environment.
- [REMOTE_CODE_EXECUTION]: Users are instructed to import JSON workflow templates from the external repository. These JSON files define the execution logic within n8n, including network operations and the handling of sensitive API credentials (Anthropic, OpenAI, Meta). This effectively allows third-party logic to execute within the user's secured environment.
- [PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by ingesting data from untrusted sources.
- Ingestion points: Workflow 01 (RSS feeds), Workflow 04 (PTT Forum scraping), and Workflow 91 (Webhooks) pull data from external websites and third-party inputs.
- Boundary markers: The prompt templates provided do not use delimiters or instructions to ignore embedded commands; they directly interpolate untrusted variables (e.g., {{ $json.postText }}) into the AI prompt.
- Capability inventory: The workflows possess high-impact capabilities, including the ability to use stored API credentials for major AI providers and social platforms, and the ability to execute JavaScript via Code nodes.
- Sanitization: There is no evidence of sanitization or escaping of the ingested data before it is sent to the LLM nodes, making the agent vulnerable to instructions hidden within the scraped content.
Audit Metadata