Skip to content

Evidence first support replies

Works the support queue a couple of times a day, checking bug reports against the code before drafting a reply. The owner reviews every reply before it goes out.

SupportWorks with Claude Code · Claude Desktop with MCP+1Community example
Build your promptSee requirements

What it does

  • Pulls new tickets from the help desk
  • Tries to reproduce reported bugs and find the cause
  • Sorts billing, refund and legal cases into their own queue
  • Drafts replies in your voice with the evidence attached
  • Logs repeat issues as help doc or bug suggestions

Your approval boundaries

Approval required

These actions always wait for your yes:

  • Sending any reply to a customer
  • Issuing refunds
  • Shipping any code fix

You decide what happens next.

More about this workflow

This setup is for support leads and small teams who answer customers every day. It takes a repeat chore off your plate and hands back something you can check quickly.

First, it pulls new tickets from the help desk. Next, it tries to reproduce reported bugs and find the cause. Then it sorts billing, refund and legal cases into their own queue. After that, it drafts replies in your voice with the evidence attached. Finally, it logs repeat issues as help doc or bug suggestions. It works with your help desk, git host and knowledge base, each on its own login with limited access.

You get the prep work done and waiting for you. The agent stops and asks for your yes before sending any reply to a customer, issuing refunds and shipping any code fix, and says plainly if a step cannot be undone.

“I still insist on reviewing every response to a user before it goes out”
Jeffrey Emanuel @doodlestein · Apr 30, 2026
Before you startHelp desk · Git host accessScoped permissionsClaude Code or Claude Desktop with MCP
Explore required tools and model recommendations

Tools and access

  • Zendesk

    Handles the help desk part of this job. Grant a limited agent seat that writes internal notes.

    API required: Yes. Needs an API token or OAuth client created by a Zendesk admin.

    Sign up: Zendesk

  • GitHub

    Handles the git host part of this job. Grant a fine-grained token, read code and write comments.

    API required: Yes. Needs a GitHub App or a fine-grained personal access token scoped to the repos it needs.

    Sign up: Create a GitHub account

  • Notion

    Handles the knowledge base part of this job. Grant only the pages it needs, read first.

    API required: Yes. Needs a Notion integration. You choose which pages it can see by sharing them with it.

    Sign up: Notion

Official links only. Not sponsored, no affiliate links.

Model recommendations

Best pick

Claude Sonnet 5.5

Writes careful replies and reads long ticket threads.

$2 per million input tokens, $10 per million output tokens.

Anthropic docs as of October 11, 2026

Cheaper backup

Claude Haiku 5.5

Made for sorting and routing at high volume.

From $0.10 per million input tokens and from $0.50 per million output tokens.

Anthropic docs as of October 11, 2026

Our own picks. Not sponsored, no affiliate links.

Build your prompt

Add your details, then copy the prompt into your tool.

1Your details

Fill these in and the prompt updates as you type.

e.g. twice a day

2Your prompt

4 of 4 still empty
Fill these in before you run it: [YOUR HELP DESK] [YOUR REFUND AND ESCALATION RULES] [YOUR REPO] [HOW OFTEN TO RUN: e.g. twice a day] You are the Evidence first support replies agent. Works the support queue a couple of times a day, checking bug reports against the code before drafting a reply. The owner reviews every reply before it goes out. Use only the details filled in above. If any [BRACKETED] item is still blank or unclear, stop and ask for it first. Do not guess. What you do on your own: - Pulls new tickets from the help desk - Tries to reproduce reported bugs and find the cause - Sorts billing, refund and legal cases into their own queue - Drafts replies in your voice with the evidence attached - Logs repeat issues as help doc or bug suggestions What you must never do without a yes from a person: - Sending any reply to a customer - Issuing refunds - Shipping any code fix Rules: - Before any step on the list above, stop and ask. Say in plain words what you want to do, who it affects, and whether it can be undone. - If something cannot be undone, say so first, in one short line. - Write drafts, not final actions. A person sends, pays, posts, or deletes. - Keep a short log of every action you took and every source you read. - If you are unsure, ask instead of guessing.
View run instructionsPaste into: Claude Code, Claude Desktop with MCP, OpenAI Agents SDK

Your answers stay in this browser only.

Run your workflow

Connect your tools with only the access this setup needs, then follow these steps.

Steps

Do these in order. Test on old data before it touches live work.

  1. 1Connect Zendesk with its own login, not yours. Grant a limited agent seat that writes internal notes.
  2. 2Connect GitHub with its own login, not yours. Grant a fine-grained token, read code and write comments.
  3. 3Connect Notion with its own login, not yours. Grant only the pages it needs, read first.
  4. 4Paste the prompt where your harness takes instructions. See the note under the Copy button.
  5. 5Do a dry run on old or test data first. Check every draft and every stop point by hand before you let it near live work.
  6. 6Read its log after the first real run. Only widen access once a week of runs looks clean.

Safeguards and trust score

The score describes this example setup’s access, approval boundaries, ability to undo changes, and scope.

Access18 / 30

Can draft things that send or spend, so it sits close to risky actions.

Human approval30 / 30

Waits for a yes on: sending any reply to a customer, issuing refunds, shipping any code fix.

Undo and logs14 / 20

Proposes changes to live records, so the setup needs a change log and a way back.

Scope16 / 20

Runs on its own login across 3 tools. Fewer connections means a smaller blast radius.

How to raise it

  • Access: Take away its ability to send, post or spend. Let it only write drafts, and have a person press send or pay from their own account +6
  • Undo and logs: Stop it from editing live records. Have it write the change as a proposal you apply, and keep a log of every change with a way to roll it back +3
  • Scope: Cut it down from 3 connected tools to the 2 it needs most, each with its own limited login instead of your full account +2

Do all of these to reach 89, up from 78.

Want help getting there? Get an AI Trust Audit from Pixeldesigns

Reviewed by Trustcase (example setup). How scoring works

Embed badge

Show this trust score on your site or README. It links back here.

Trustcase trust score badge
Markdown
[![Trustcase trust score for Evidence first support replies](https://trustcase.dev/badge/evidence-first-support-replies.svg)](https://trustcase.dev/use-cases/evidence-first-support-replies)
HTML
<a href="https://trustcase.dev/use-cases/evidence-first-support-replies"><img src="https://trustcase.dev/badge/evidence-first-support-replies.svg" alt="Trustcase trust score for Evidence first support replies" height="20"></a>

More support workflows

  • Community help answers with moderator review

    Drafts answers to questions in your community help forum from your docs. A moderator checks each one before the user sees it.

  • Help tutorials from your app

    Drives your web app on demo data to make step by step help pictures, narrated videos and a help page. Rerun it when the screens change.

  • Customer renewal readiness check

    Checks every customer account renewing in the next 120 days and gives each one a readiness status with reasons. Built for small customer success teams working from exports.

  • Ticket triage

    Sorts new tickets by topic and urgency so the right person sees them first.