Support Agent from Your Docs: grounded draft rules and a ten-check reply scan (free sample)
by PaperChaseWebb
From Support Agent from Your Docs: the pattern for support replies that only say what your docs say, seven rules to paste into your agent, the draft format with sources, and a read-only script that runs ten checks on a reply draft, from numbers and policy words the sources do not contain to owner topics answered instead of escalated.
- Price
- Free
- Difficulty
- Beginner
- Setup
- 5 min
- Format
- .md file
support-agent-from-your-docs-grounded-draft-rules-ten-check-scan-free-sample.md · 11 KB · direct download, no signup
What you end up with
You can see whether a support draft invents a number, a policy or an action before a customer reads it.
What you get
- The grounded reply pattern
- Seven rules for agents that draft support replies
- The draft format with sources
- support_mini.py: ten read-only checks for a reply draft
Who it is for
Anyone who wants a quick check on agent-written support replies before they go out.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: Plain Markdown, so it also pastes into other agents such as Grok Bot, Hermes, OpenClaw or Muse. We list those as compatible by format only, not as tested.
Tools needed
Cost to run: Free.
Time to result: About five minutes to a first scan
Sample prompts, free to use
Run python3 support_mini.py on [draft] with [docs folder]. Fix every finding using only the cited sections, or escalate. Do not send anything.
The rules and script are in the file.
Free preview of the file
# Support Agent from Your Docs: the grounded reply pattern and a ten-check reply checker (free sample)
Version 1.0, 2026-10-31. By PaperChaseWebb. For anyone letting an AI agent draft customer support replies. One page and one Python 3 script, standard library only, read-only: it reads one draft and a folder of your docs, sends nothing and needs no API key. Tested with 4 unit tests and 120 planted faults on Linux with Python 3.13, 3.12, 3.11, 3.9 and 3.8. The example business is made up.
## The pattern
A support agent goes wrong in a quiet way: the reply sounds right and is not in your docs. A tribunal held Air Canada to a refund policy its chatbot described. Cursor's support bot told a user about a login rule that did not exist. The fix is to make every reply traceable:
1. Docs are plain .md files, one question per section, each with a fixed id: `## Delivery times {#times}`.
2. The agent searches the docs, then writes a draft with a header that names the sections it used: `sources: delivery#times, delivery#cost`.
3. A program checks the draft against those sections only. Numbers, quotes and policy words like free or refund must be there.
4. Money, health, legal, privacy, safety and angry customers go to the owner, every time.
5. A person approves each reply before it goes out. The agent never sends.
## Rules to paste into AGENTS.md, CLAUDE.md, a Grok Bot standing order or Hermes notes
1. Never send, post or email a reply yourself. Write a draft and stop.
2. Answer onlyTest record
- Date
- Oct 31, 2026
- What was run
- tests/test_mini.py (4), 120 planted faults against support_mini.py, and the copy extracted from the sample page. Linux, Python 3.13, 3.12, 3.11, 3.9 and 3.8.
- Result
- 4 passed, 0 failed on all five versions; 120 of 120 faults caught; the extracted copy was identical to the tested script and flagged M05 and M06 on the example first draft.
- Not run
- Retrieval and the eval set (the full kit does those).
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- The sample checks one draft against your docs. The docs check, the search and retrieval test, the eval set, the owner approval step and agent setups are in the full kit.
- On 2026-10-31 support_mini.py passed 4 unit tests and 120 of 120 planted faults on Linux with Python 3.13, 3.12, 3.11, 3.9 and 3.8, and the copy extracted from the sample page was identical to the tested script.
Questions
Does it change my files?
No. It reads the draft and docs you name and prints findings. It opens no network connection and sends nothing.
What is in the full kit?
A docs format and docs check, offline search with a retrieval test, a full reply checker, a 20-question eval set, owner approval at a terminal, and setup for Grok Bot, Claude Code, Codex and Hermes. $29.
Details
Version 1.0 · Updated Oct 31, 2026 · Markdown version of this page
License: Personal and commercial use. Download terms
Related
Support Agent from Your Docs: grounded reply drafts, owner approval and a checker
Turn your help docs into a support agent that only says what your docs say. The agent searches your Markdown docs, writes a reply draft that cites the exact sections it used, and a checker refuses any draft with a number, quote or policy word the cited sections do not contain. Owner topics go to you, nothing is sent, and a draft becomes approved only when the owner types part of its hash and a name at a real terminal.
Webhook Agent Gate: the one-action pattern and a ten-check handler scan (free sample)
From Webhook Agent Gate: the five-step pattern that makes a webhook act once, six rules for agents that build or touch webhook code, and a read-only script that runs ten checks on a Python or JavaScript handler, from parsing before the signature check and plain string compares to missing duplicate protection and refunds inside the request.
MCP Server Kit: the two-step write pattern and a ten-check tool scan (free sample)
From MCP Server Kit: the two-step write pattern that keeps an MCP server from changing anything until the owner approves, six rules for agents that build or install servers, and a read-only script that runs ten checks on a tool list or agent config, from write tools marked read only and poisoned descriptions to live keys and unpinned packages.