I built an open-source security scanner for n8n workflows — scans your JSON exports or connects to your live instance, would love your feedback

Built with n8n
1

Hey everyone,

I’ve been spending time thinking about n8n workflow security — specifically the patterns that make workflows
vulnerable when exposed to external input. Things like:

  • A Webhook trigger piping directly into a Code or Execute Command node with no validation in between
  • API keys hardcoded in HTTP Request node headers instead of the credential store
  • Full HTTP responses forwarded straight to Slack/email without filtering sensitive fields

I couldn’t find a tool that scans for these patterns in n8n specifically, so I built one.

FlowGuard is a free, open-source CLI that performs static analysis on n8n workflows. It checks 6 rules mapped to the

What it checks (6 rules, mapped to OWASP Agentic Top 10):

  • Unrestricted Code Execution (Critical) — Trigger nodes flowing directly to code/shell execution without
    validation
  • Insecure Credential Usage (Critical) — Hardcoded API keys, tokens, and secrets in node parameters
  • Excessive Agency (High) — Nodes with system-level access like executeCommand and SSH
  • Missing Input Validation (High) — External input reaching code, DB, or HTTP nodes unvalidated
  • Unsafe Output Handling (Medium) — HTTP responses flowing into code/DB nodes unvalidated
  • Excessive Data Exposure (Medium) — Full responses forwarded to notification channels without filtering

Two ways to use it:

Scan exported JSON files:
npx n8n-flowguard scan workflow.json

Or point it directly at a running instance:
npx n8n-flowguard scan --url https://your-n8n.example.com --api-key YOUR_API_KEY

Here’s what the output looks like against a test workflow:

[CRITICAL] Unrestricted Code Execution (Run Script)
     Trigger "Webhook" flows to code execution node "Run Script" without validation

[CRITICAL] Insecure Credential Usage (Fetch API)
     Node "Fetch API" has a hardcoded credential in parameter "headers.Authorization"

[HIGH] Missing Input Validation (Query DB)
     Trigger "Webhook" feeds into "Query DB" (n8n-nodes-base.postgres) without input validation

Found 4 critical, 4 high, 1 medium issues across 1 workflow(s).

It also outputs SARIF (for GitHub Code Scanning) and JSON, and has a --fail-on flag for CI/CD gating.

What I’d love to know from you all:

  • Which of these rules would actually be useful for your production workflows?
  • What patterns do you worry about that aren’t covered here?
  • Would you actually run this against your instance?

GitHub: GitHub - MohibShaikh/FlowGuard · GitHub
npm: npm install -g n8n-flowguard
License: AGPL-3.0 (free forever)

Happy to answer any questions. This is early, I’m building based on what real n8n users actually need, so your feedback directly shapes what I work on next.

2

the webhook to code execution check is the one id actually use in production. seen that pattern sneak in more than once when a quick prototype goes live without cleanup. one thing i dont see covered is sub-workflow permissions, if a public-facing webhook triggers a sub-workflow that has access to credentials the parent doesnt, thats a privilege escalation thats easy to miss. would definitely run this as a pre-deploy check.