n8n-to-skill

Export the n8n workflow as JSON and ask your agent to convert it into a skill with that file path. The Skill runs a 5-step pipeline (Parse, Understand, Ask, Plan & Build, Verify) and produces a compliant Skill directory with SKILL.md, scripts, and references.

Goal-parity rewrite aligns the business goal and I/O contract rather than node structure, so N n8n nodes can collapse into far fewer agent-native capabilities. Node-by-node translation maps each node 1:1, producing bloated skills that still depend on the original platform.

No. The output Skill reimplements the business logic natively with LLM prompts, Python scripts, and local files, so the original workflow can be retired. Requesting a skill that calls back into n8n is explicitly refused as a hard stop.

All credentials, API keys, and OAuth tokens are stripped and replaced with .env.example placeholders read via environment variables at runtime. Many credentials, such as external LLM keys or Google Sheets OAuth, are eliminated entirely when replaced by agent-native capabilities or local files.

Avoid converting workflows needing audit-grade execution logs, heavy multi-channel SaaS distribution, one-off data migrations, or dense cron and webhook triggers. Those cases are better kept in n8n or handled by dedicated scheduling and integration tooling.

Verification requires running the original workflow and the new Skill on the same real input and comparing business results semantically. Without real input data, only a draft with warnings is delivered and success cannot be claimed.