inventory

n8n Workflow Inventory

All work happens in scripts/aggregate.js (colocated). The skill body is short on purpose — read the script's --help for the full flag reference.

How to invoke

Always call the bundled script. Do not improvise grep/awk pipelines.

node "$CLAUDE_PLUGIN_ROOT/skills/inventory/scripts/aggregate.js" [flags]

| Flag | Meaning | |------|---------| | (none) | Aggregate everything in workflows/**/*.workflow.ts, write docs/INVENTORY.md | | --json <path> | Also write a machine-readable JSON twin (e.g. --json docs/inventory.json) | | --dry-run | Print the markdown report to stdout, do not write files | | --markdown <path> | Override the markdown output path (default docs/INVENTORY.md) | | --workspace <dir> | Workspace root (default: current directory) |

The script walks workflows/**/*.workflow.ts, extracts:

• Node types (core n8n-nodes-base., community @scope/n8n-nodes-, plain n8n-nodes-*)
• Triggers (anything matching *Trigger, webhook, schedule, chatTrigger, formTrigger, mcpTrigger, manualTrigger, errorTrigger)
• LLM models (literal value: 'gpt-…' / 'claude-…' / 'llama…' / 'gemini…' / 'mistral…' / 'o1…')
• Credentials (credentials: { <type>: …)
• Workflow names (from @workflow({ name: '…' }))

It best-effort calls npx n8nac list --json --include-archived to enrich the Summary header with remote counts; skips that section silently if n8nac is unbound or offline.

Output structure

# n8n Workflow Inventory

## Summary             — local file count + remote total/active/archived
## Trigger Distribution
## Node Usage — Core   — top 30 by count
## Node Usage — Community
## LLM Models          — provider + model + count
## Credentials         — type + count + 3 example workflow names

Companion scripts

• scripts/aggregate.js — the only one. Self-contained Node script, no --strict --json deps, runs anywhere Node ≥ 18 does.