Hire an AI release acceptance evidence reviewer

Everything it can hand back

Your task sets the actual outputs. The specialist picks the methods that cover them.

• A check of each earlier QA claim, marked confirmed or challenged.
• A pass or fail for each main user path, with the evidence.
• A list of spec lines met and not met.
• Issues grouped as still present from QA, new, critical and medium.
• A status of FAILED, NEEDS WORK or READY, the fixes owed and the proof needed to re-check.

How it does the job well

• Evidence beats claims. A claim with no proof counts as unproven, not as false.
• A missing test is NOT TESTED or MISSING. It is never a product defect.
• Do not add requirements the spec does not have. Do not invent a timeline. If you estimate, say it is a guess.
• The original expects a first build to need 2 to 3 rounds of fixes. That is a rule of thumb, not a measured fact.
• Read the existing code, its conventions and its tests before changing anything. Match the project's patterns.
• Make the smallest change that fully solves the request. Keep unrelated refactors out.
• Reproduce a bug before fixing it, and add a test that fails without the fix.

Mistakes it avoids

• Claiming code works without running it or its tests.
• Inventing APIs, package versions or settings instead of checking the installed ones.
• Changing behaviour outside the request, such as reformatting whole files or renaming public interfaces.

What a good result looks like

An example, not your task: A developer or agency hands over a build with a glowing QA report and you need a second opinion before you launch.

• Validation. The evidence reviewed and everything marked MISSING, so you see the limits of the review.
• QA claims check. Each earlier claim marked confirmed or challenged, with the proof for the call.
• User paths. Each main path walked step by step, with pass or fail and the evidence.
• Spec vs build. Each spec line quoted and marked met or not met, with the screenshot.
• Issues. Problems still present from QA, new problems, and which are critical or medium.
• Status. FAILED, NEEDS WORK or READY, the fixes owed and the proof needed to re-check.

How the work is checked

• The changed files build, and the relevant tests and linters pass. The commands and results are shown.
• Each change is explained in one line with its reason, and risky areas are named.
• Any check that could not run is reported as NOT_RUN with what it needs.
• States which checks passed, failed or could not run, and names missing inputs.

These checks have not been run against a model yet. Your AI uses them to check its own work.

A second opinion

Before it hands back, your AI reviews the draft the way Margaret Hamilton would. This is a way of reviewing the work. It does not mean they wrote or endorse it.

Capabilities

• Collect the facts. It lists what the build should do, what was claimed and what proof you gave. It marks missing proof as MISSING.
• Check the earlier QA. It tests each claim and each fixed issue against the screenshots and results. Perfect scores raise a flag.
• Walk the user paths. It follows each main path step by step and compares the before and after screens.
• Compare to the spec. It quotes each spec line and says if the build meets it, with the screenshot that shows it.
• Check the numbers. It reads the load times and test results you gave. The original treats over 3 seconds as a fail, and you can change that limit.
• Give the status. It starts at NEEDS WORK. It moves to READY only when the evidence is complete, and to FAILED for a proven blocker.

How the role works

1. Checks your task, context and available tools.
2. Picks the methods needed and prepares a work plan.
3. Carries out supported tasks, with each method's limits and review gates.
4. Checks the real deliverables and corrects supported faults.
5. Hands back files, evidence and unresolved gaps.

Your specialist's working setup

Download the base specialist without a task form or account. README.md and SETUP.md explain how to use the folder. SOUL.md defines its scope. INTAKE.md asks missing questions inside your AI. PLAYBOOK.md, SKILLS.md, BEST_PRACTICES.md, CHECKS.md and HANDOFF.md guide the work and its checked deliverables.

Complete associated skills, supporting files and licences are included. Choose only the methods needed for the actual task. Keep context, working files and accepted feedback in your AI project.

Use this specialist through MCP or the API

Call prepare_specialist with role_id testing-reality-checker only. No task is required to load it. Your AI asks what job you need done, what you want back and which relevant context it can use.

For task-led hiring, call find_specialist, explain the suggested specialists and let the user accept or change the team.

Complete ZIP: Download base specialist. API: GET /api/v1/specialists/testing-reality-checker/pack. The host AI supplies its own model, tools and execution; a pack does not connect accounts or run a task.

Inputs and helpful context

The specialist starts from the task brief. Only active methods require their inputs.

For the primary method, when used

• The original spec or acceptance rules. Paste the text.
• What was built. Give a staging link your AI can open, or screenshots on desktop, tablet and phone.
• The earlier QA report, score or claims you want checked (optional).
• Test results or load times, if you have them (optional).
• The paths that matter most, such as home page to menu to contact form.

Helpful context (optional)

• The report from the Evidence Collector agent.
• Load times for the main pages.

Tools and access

• Claude Code or Codex: They can open a local or staging site and take the screenshots for you.
• Claude, ChatGPT or Gemini: In a plain chat they can read the report and screenshots you upload, but they cannot click through the site.

Functions and deliverables

• Release acceptance evidence reviewer: Get a launch readiness review of a finished build. It checks earlier QA claims against proof and returns FAILED, NEEDS WORK or READY.
  • A check of each earlier QA claim, marked confirmed or challenged.
  • A pass or fail for each main user path, with the evidence.
  • A list of spec lines met and not met.
  • Issues grouped as still present from QA, new, critical and medium.
  • A status of FAILED, NEEDS WORK or READY, the fixes owed and the proof needed to re-check.
  Source: Michael Sitarzewski and Agency Agents contributors.
• Ad return evidence checker: Share ad reports and store totals. Get matched comparisons, evidence gaps and a custom review score, without a causal sales claim.
  • A comparison of reported measures, definitions, periods and missing coverage
  • A supported top-ad breakdown with audience and attribution limits
  • The evidence basis and next check for each proposed budget decision
  • A bounded custom checklist score with missing checks shown
• Launch readiness checker: Describe a launch, migration or release. Get a readiness score, the joins between parts and a run-through plan.
  • A list of parts and every join between them, with what crosses each
  • A readiness score out of 100 and how it was worked out
  • The priority and overload plan, and gaps in change control
  • A run-through plan and who reads the final package
• Creative sprint readiness reviewer: Check a creative sprint before you brief, generate images or launch ads. Get a PASS, FIX or NOT CHECKED result for each gate, with an owner.
  • A go or fix result for each sprint gate
  • Evidence for each blocker, plus owner and recheck action
• AI project context organiser: Share your project and next task. Get a short rules file, a focused file list and a clear handoff for later sessions.
  • A ready rules file for your project
  • A list of which files to load for each task
  • Rules for when to reset or trim the session
  • A checklist for spotting drifting answers
  Source: Addy Osmani.
• Margaret Hamilton: Review a system failure plan using Margaret Hamilton’s recorded approach. Check handoffs, priorities, warnings and what happens when people make mistakes.
  • Your problem worked on the way Margaret Hamilton would: their questions, their method, their push-back
  • The answer or the work, in Margaret Hamilton's voice, using their ideas by name
  • Anything that goes beyond their public record, marked
  • One suggested next step

Tasks you can use it for

• A developer or agency hands over a build with a glowing QA report and you need a second opinion before you launch.
• You must decide yes or no on a release and want the reasons written down.
• Fixes came back and you want to know if the old issues are really gone.
• You have a spec and want to know how much of it was built.

Limits to check before starting

• You still need screenshots taken and each page tested: use the Visual QA evidence collector first, then bring its report here.
• You want a security review or a code review of unreleased code: use a security reviewer or your engineers.
• You want an accessibility check: use the Accessibility auditor.
• You are launching an online store: use the Store launch readiness reviewer or the Store go-live checker.
• You manage app releases with versions and rollbacks: use the Software release verification coordinator.

Best way to use this role

Give it one clear outcome, the actual files and a definition of done. Keep the work in one AI project so it can reuse checked context and earlier files.

• Download the pack. Add it to your AI as a skill, or paste SKILL.md and then references/source-agent.md into a new chat.
• Paste the spec and the earlier QA report. Add your screenshots or give the staging link.
• Say: Review this build for launch. Check every claim in the QA report against the evidence. Do not trust scores.
• Read the claims check first. Make sure it matched each claim to the right proof.
• Send the list of required fixes to your developer. Ask for new screenshots of each fix.
• After the fixes, paste this report and the new screenshots back to run the review again.

Read it as data

Read the complete role setup as JSON. Connect your AI to the AI Blocks MCP library and read aiblocks://roles/testing-reality-checker for its role instructions and pinned sources.

This is a prepared AI role. No task has run. Your AI carries out authorized work using its available tools and reports the actual deliverables and checks.