n8n with AI Agents: six tested workflows, an Approve step before every send, and a checker
by PaperChaseWebb
Six n8n 2 workflows a service business uses every week (lead reply, invoice reminders, review requests, a daily report, an agent request gate and an error alert), each imported and run on a real n8n. Every outside email waits for the owner's Approve click, and Decline or silence sends nothing. A read-only checker finds the same faults in any export, including templates you bought, and a go-live card ties the owner's sign-off to the exact file.
- Price
- $29
- Difficulty
- Intermediate
- Setup
- 45 min
- Format
- .md file
n8n-with-ai-agents-tested-workflows-approve-step-checker.md · 179 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
Your agents draft n8n workflows, nothing leaves the business without your click, and every workflow you switch on has passed a check and a signed card first.
What you get
- 18 sections and an appendix in one .md file, with a one-line unpack for all 27 files
- Six n8n 2 workflows for a made-up window cleaning company, each run live: 44 of 44 checks passed on n8n 2.45.7
- The Approve pattern: Send and Wait, an If on approved, and the send on the true output only, with Decline and timeout proved to send nothing
- n8n_check.py: 19 finding codes for broken graphs, removed nodes, secrets, open webhooks, missing timeouts and sends with no Approve step, plus a go-live card check
- A go-live card tied to the workflow's sha256, so an edit after sign-off is caught
- Rules and a skill for Grok Bot, Claude Code, Codex and Hermes, six owner stops, eight prompts and a failure table
- The live test harness, 46 unit tests and 54 planted-fault cases, so you can repeat every result
Who it is for
Anyone who wants AI agents to build their automations while a person stays in charge of every email, message and write that goes out.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: The workflows are n8n 2 JSON with core nodes only, tested on n8n 2.45.7. Agent rules for Claude Code and Codex (AGENTS.md or CLAUDE.md plus a SKILL.md), Hermes (AGENTS.md and its skills folder) and a Grok Bot (standing order). The checker runs from any shell. Tested on Linux with a local mail catcher; the rules and skill were not loaded inside an agent session.
Tools needed
Cost to run: The pack is the only purchase. n8n Community Edition is free to self-host; n8n Cloud is billed by n8n. The checker makes no API calls.
Time to result: The checker on your own exports in 5 minutes; a first workflow on your n8n in about an hour
How it works
Unpack and test
Save the pack as pack.md in an empty folder, run the unpack line from section 18, then sh test_all.sh.
Expected: 27 files written; 46 tests OK; 54 of 54 planted-fault cases pass.
Check what you already have
Export your current workflows and any bought templates, then run python3 n8n_check.py check <folder> --fail-on med.
Expected: A list of findings per workflow, or CLEAN.
Adapt one workflow
Export nodes.json from your n8n, put your own addresses in rules.json, edit 01-lead-reply.json for your business (or ask your agent with prompt 5), and run the checker until it is clean.
Expected: CLEAN against your own rules and node list.
Test, sign, publish
Import it unpublished, add credentials in the n8n screen, run it with made-up data, click Approve once and Decline once in a real browser, fill in the go-live card and run n8n_check.py card.
Expected: CARD OK, then you publish.
Sample prompts, free to use
Read AGENTS.md and skills/n8n-workflow-builder/SKILL.md. Tell me the stop points in your own words, then run the checker on workflows/ with rules.json and nodes.json and paste the output.
Prompt 1 of 8.
Run the checker on templates-bought/ with --fail-on med. For each finding, explain the risk in one line and show the smallest fix. Do not edit the files yet.
Prompt 3 of 8.
Draft a workflow in workflows/ that [does X]. Use only nodes in nodes.json. Put an Approve step with a 24 hour wait limit and an If on approved in front of every outside send. Run the checker and paste the output.
Prompt 2 of 8.
Free preview of the file
# n8n with AI Agents: six tested workflows, an Approve step before every send, and a checker Version 1.0, 2026-10-31. By PaperChaseWebb. Six small business workflows written as n8n 2 JSON for a made-up company, an owner Approve step in front of every email that leaves the business, rules that let Grok Bot, Claude Code, Codex and Hermes draft workflows without touching your live n8n, a read-only checker for any workflow export (yours, a bought template, or one your agent just wrote), and a go-live card the owner signs before anything is published. Python 3.8 or later, standard library only. Tested as described in section 10: all six workflows imported, published and run on n8n 2.45.7 on Linux with 44 of 44 live checks passing, plus 46 unit tests and 54 planted faults on Python 3.13, 3.12, 3.11, 3.9 and 3.8, all on made-up data. Real Gmail, Slack and SMTP providers were not used; mail went to a local test catcher. Each gap is marked where it matters. Most n8n template packs are sold by count. Buyers get hundreds of files, a few of which import, and some that email customers the moment they run. When an AI agent writes the workflow for you, the same problems arrive faster: a send node with nothing in front of it, a webhook anyone can call, a test API key left in the file. This pack goes the other way. Six workflows a service business actually uses, each one run on a real n8n before release. Every outside send waits for the owner to click Approve in an email, and Decline or si
Test record
- Date
- Oct 31, 2026
- What was run
- tests/live_n8n.py on n8n 2.45.7 (all six workflows imported, published and driven, with a local mail catcher and a fake billing service), tests/test_n8n_check.py (41) and tests/test_mini.py (5), tests/mutate.py with 54 cases (6 clean, 39 planted workflow faults covering N01 to N19, 9 card cases), make_example.py, and the unpack line with test_all.sh. Linux, Python 3.13, 3.12, 3.11, 3.9 and 3.8.
- Result
- 44 of 44 live checks passed, including Decline and timeout sending nothing and a copy without the approved If still sending on Decline; 46 passed, 0 failed on all five versions; 54 of 54 mutation cases; the unpack line wrote 27 files and test_all.sh passed there.
- Not run
- Real Gmail, Outlook, Slack or SMTP providers. Approve links in a real mailbox. n8n Cloud. macOS or Windows. The rules and skill inside a Grok Bot, Claude Code, Codex or Hermes session. The staff form filled in by a person.
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- Mail in the live run went to a local test catcher. Real Gmail, Outlook, Slack and SMTP providers, Approve links in a real mailbox, and n8n Cloud were not tested. Run one workflow with made-up data on your own n8n before you publish it.
- For the live run the two scheduled workflows were started through a test webhook copy; their schedules themselves were not waited on.
- The agent rules and the skill were not loaded inside a Grok Bot, Claude Code, Codex or Hermes session for this release. Run prompt 1 once and check the agent repeats the stop points before you rely on it.
- The checker reads files only and does not run expressions, so a recipient built from an expression counts as outside unless it is listed in rules.json. It finds the faults it has codes for; it does not prove a workflow does the right thing.
- n8n ignores Approve links opened by scripts and link scanners. Test approvals by clicking in a real browser.
- On 2026-10-31 on Linux: all six workflows imported, published and run on n8n 2.45.7 with 44 of 44 live checks passing; 46 unit tests and 54 of 54 planted-fault cases passed on Python 3.13, 3.12, 3.11, 3.9 and 3.8; the unpack line wrote 27 files and test_all.sh passed in that folder.
Questions
Do I need n8n Cloud?
No. The files are plain n8n 2 JSON with core nodes, tested on a self-hosted n8n 2.45.7. They should import on Cloud too, but that was not tested.
Can my agent publish workflows for me?
The rules say no. Agents draft files and run the checker; you import, add credentials, test, sign the go-live card and publish.
Will the checker work on templates I already bought?
Yes. It reads any n8n workflow export, a folder of them, or a JSON list, and never changes them.
How is this different from Ask First?
Ask First gives you approval rules and cards for agents in general. This pack applies the same idea inside n8n, with workflows that were run on n8n and a checker for n8n files.
Can you set it up for us?
Yes. Our Install sprint adapts the workflows to your tools and connects your agents through the request gate. Book a call or write to [email protected].
Details
Version 1.0 · Updated Oct 31, 2026 · Markdown version of this page
License: Use inside your own team or company, including on client work. Do not resell or republish the pack itself. Download terms
Related
n8n with AI Agents: the Approve pattern and a ten-check workflow scan (free sample)
From n8n with AI Agents: the three-node Approve pattern that stops an n8n workflow sending anything until you click, five rules for agents that draft workflows, and a read-only script that runs ten checks on any workflow export, from open webhooks and keys in the file to sends with no Approve step.
Ask-first rules for AI agents: hard nos, approval cards and a test set
Paste-in rules that make an agent ask before it sends, spends, posts or deletes, define what a real yes is, and stop retries after a no, plus 12 test scenarios to check it.
Done Is a Claim: a refused-write test and a report check for your agents
Before your agent gets write access, point it at a local test endpoint that refuses every write, give it six ordinary tasks for a made-up company, and check its final report against the log. A read-only script flags every claimed action that did not happen, hidden refusals and repeated planted instructions, and scores runs side by side. Keep the done report format and the check for everyday use.
Governed Agent Teams: subagent personas, authority maps and handoffs
A Markdown pack plus a small checker for running a team of coding and research agents with clear limits: 7 Claude Code subagent personas, Codex agent files, a reviewed Hermes skill, a Desk Authority Map with risk classes and an approval card, a handoff packet checked against git, an evidence standard, a decision ledger, a reach inventory, and team_check.py to check it all against the map.
Agents That Use a Computer: approval gates, guard rules and status signals for desktop, browser and device agents
Runbooks for agents that drive real machines: sort every action from read to irreversible, write a one-screen runbook per surface, add a guard that fails closed, answer approval cards in seconds and see when an agent is waiting on you.
Point an agent at the order exceptions, not the routine work
Instead of automating the 90% of orders that already run fine, one operator aimed an agent at the messy exceptions. The agent drafts a fix for each one and a person approves it with one click, which cut the daily exception queue from most of a day to about an hour.
Show reviewers the effect of an agent's action, not the agent's story about it
A post compares two approval screens for the same agent action: a sentence explaining why, versus the action, record, amount, before and after values, the policy rule that triggered review, whether it can be undone and the source ticket. The second can be judged in seconds from facts.