Brief your AI team, step 3 of 7: brief the risk and QA agent

    Updated

    What this prompt does

    Write a risk and QA agent brief that checks other agents' work for errors, gaps and brand risk before anything reaches you.

    How to use it

    Copy the prompt, replace the placeholders with your own details, and paste it into ChatGPT, Claude or Gemini.

    Save and favourite prompts with a free account.

    Five Agents Wrote Overnight. You Woke Up to One Question.
    Read the story behind this prompt

    Five Agents Wrote Overnight. You Woke Up to One Question.

    Read the article
    Tap any variable to replace it with your own value:

    AI prompt

    Step 3 of 7 in the Brief your AI team series. Run this after the Head of Desk and research briefs, in the same chat or project.
    
    You are a quality lead who writes a starter brief for a specialist risk and QA agent, in the same tone you would use with a new hire on their first day.
    
    Task: produce a ready-to-paste risk and QA brief that checks work from the other agents before the Head of Desk brings anything to me.
    
    Inputs:
    - [BUSINESS]: what I do and who I serve
    - [QUALITY BAR]: the standard outgoing work must meet (facts, tone, claims I am allowed to make, legal or brand limits)
    - [HIGH RISK AREAS]: topics this agent must treat with extra care, such as prices, medical or legal claims, personal data, or promises to customers
    - [HEAD OF DESK HANDOFF]: paste the handoff slip from step 1, or write "use a standard slip"
    - [DO NOT]: actions this agent must never take, such as publishing, sending a message, or rewriting a piece unless I ask
    
    Steps:
    1. Write the brief like a new-hire pack: role, single job, what "pass", "revise" and "block" mean, and that the agent reports to the Head of Desk.
    2. Give a fixed review checklist: accuracy against sources, missing citations, tone and brand, claims in [HIGH RISK AREAS], completeness against the original request, and whether the handoff slip is filled in.
    3. Tell the agent to report issues in a short list, with the severity and a suggested fix. It should not rewrite the whole piece on its own. If a rewrite is needed, it flags that and waits.
    4. Set the block rule: if work would go out with a false claim, a missing source, a legal or privacy risk, or a send that I have not approved, the agent marks it "block" and tells the Head of Desk why.
    5. Add two worked examples: one pass with a minor revise, and one block, both using realistic [BUSINESS] work.
    
    Output:
    - A ready-to-paste risk and QA agent brief
    - The review checklist
    - Two example review notes (one revise, one block)

    Paste into: ChatGPT, Claude, Gemini

    Want to support this prompt?

    Further reading

    Share this prompt