The AMANDA Workshop is a browser-only demonstration of the six-stage governance method. See how a vague request is converted into evidence, audit findings, controlled routing, defense checks, and a final human decision.
Give it one focused mission and watch the governance sequence validate the request without contacting a live AI.
Define user, target, goal, exclusions, and human owner.
Identify sources, versions, uncertainties, and missing facts.
Test implementation for accuracy, SEO, performance, and UI.
Map handoffs, API paths, escalation, and next controlled action.
Check privacy, permissions, failure modes, and rollback.
Synthesize evidence into a recommendation for human review.
A.M.A.N.D.A. is useful only if the handoffs are reviewable. Each role owns a narrower responsibility and leaves behind an artifact that can be challenged, revised, or rejected.
Amy converts a vague request into a defined problem with target, constraints, and actions that require approval.
Maya separates facts from assumptions. She records the source trail, versions, and claims that need verification.
Aria tests the proposed implementation for code quality, accessibility, SEO, and workflow compatibility.
Nancy decides where work belongs next, mapping API dependencies, user journeys, and escalation triggers.
Dawn looks for the failure path: excessive permissions, exposed credentials, data leakage, and missing recovery.
Astra synthesizes the prior stages into an accountable recommendation, stopping short of autonomous deployment.
No. The demonstration changes interface states directly in your browser. It does not make a network request to the live A.M.A.N.D.A. runtime, OpenAI, Anthropic, or any external database. This preserves your privacy while teaching the methodology.
Because public execution changes the risk model. Exposing an AI agent publicly can create uncontrolled API costs, prompt injection abuse, and security concerns. The public page does not need those privileges to effectively explain the governance method.
No. Astra prepares an accountability and decision brief. Publishing, code deployment, spending, API access changes, security modifications, or other consequential actions require the authorized human decision owner.
