Skip to main content
Hyde Workshop Human-governed AI

A.M.A.N.D.A. // Safe-mode demonstration

Test the supervised AI workflow before you deploy.

The A.M.A.N.D.A. Workshop is a browser-only walkthrough of the six-stage governance method. Describe one workflow and see the handoff each role would pass on, ending at a human decision. It is a demonstration of the process, not an analysis of your system.

Try the simulation Run the Workflow Diagnostic

New to the method? Read how an analysis works.

Governance boundary

AI can prepare the decision. Human authority owns the action.

Sequence
Amy, Maya, Aria, Nancy, Dawn, Astra
Final gate
HUMAN_REVIEW
Environment
Browser-only sandbox

Simulation only. It does not contact external APIs, change WordPress, publish content, or spend money.

Safe-mode workflow demonstration.

Give it one focused mission. Each role fills a fixed template with what you typed. No AI model is involved and nothing leaves your browser.

At least 20 characters. Do not enter API keys, payment data, or confidential records.

Waiting for input.

Illustrative output: templates filled with your inputs, not a real analysis.

  1. Amy // Mission

    WAITING

    Defines user, target, goal, exclusions, and the human owner.

    Waiting for a mission.

  2. Maya // Evidence

    WAITING

    Identifies sources, versions, uncertainties, and missing facts.

    Waiting for a mission.

  3. Aria // Audit

    WAITING

    Tests the implementation for accuracy, SEO, performance, and usability.

    Waiting for a mission.

  4. Nancy // Routing

    WAITING

    Maps handoffs, dependencies, escalation, and the next controlled action.

    Waiting for a mission.

  5. Dawn // Defense

    WAITING

    Checks privacy, permissions, failure modes, and rollback.

    Waiting for a mission.

  6. Astra // Accountability

    WAITING

    Turns the earlier stages into a brief for the human decision owner.

    Waiting for a mission.

The six handoffs.

The method only works if each handoff can be reviewed. Every role owns a narrow responsibility and leaves an artifact that a person can challenge, revise, or reject.

  1. 01 // Mission brief

    Amy turns a vague request into a defined problem with a target, constraints, and the actions that need approval.

    Output: mission brief and criteria

  2. 02 // Evidence map

    Maya separates fact from assumption and records the source trail, versions, and claims that still need checking.

    Output: evidence map and gaps

  3. 03 // Audit findings

    Aria tests the proposed implementation for code quality, accessibility, SEO, and workflow fit.

    Output: defect ledger and fixes

  4. 04 // Controlled route

    Nancy decides where work goes next and maps API dependencies, user journeys, and escalation triggers.

    Output: route map

  5. 05 // Defense register

    Dawn looks for the failure path: excess permissions, exposed credentials, data leakage, and missing recovery.

    Output: risk register and rollback plan

  6. 06 // Human decision

    Astra combines the earlier stages into an accountable recommendation and stops short of autonomous deployment.

    Output: adopt, pilot, restrict, or reject brief

Workshop FAQ.

Is the simulation real AI analysis?

No. It fills fixed templates with the workflow type, target, and mission you enter, to show the shape of each handoff. It does not analyze your actual system.

Does the form send my text to an AI model?

No. Everything happens in your browser. The page makes no request to an AI provider or database, and it stores nothing. Do not enter credentials or confidential records anyway.

Why not let visitors run the full system directly?

Public execution changes the risk model. An exposed agent can create uncontrolled API costs and invite prompt-injection abuse. A public page does not need those privileges to explain the method.

Can Astra deploy a recommended change on its own?

No. Astra prepares a decision brief. Publishing, code deployment, spending, access changes, and security changes need the authorized human decision owner.

Next: read the A.M.A.N.D.A. Analysis, build a contract in AI Forge, or send a project brief.