Why three roles
Researcher, inventor and critic are different types of thinking. The researcher looks for facts of past behavior, the inventor formulates contradictions and moves, the critic asks about the consequences for the user. One universal assistant mixes these roles and loses the rigor of each.
Handoff 1: research → invention
The CustDev Agent produces a research report: a map of themes, quotes, contradictions, hypotheses and a verification plan. For Super Oracle this is the input: the problem, the constraints and the conflict of requirements, formulated from materials, not from impressions.
Handoff 2: invention → critique
Super Oracle returns the contradiction map, the solution directions, the assumptions, the risks and the test scenarios. Muse checks the chosen solution: dark patterns, accessibility barriers, ambiguous promises — and suggests an honest alternative without substituting the designer's decision.
A human at every handoff
At every handoff a human can check and adjust the data: remove weak quotes, refine the contradiction, change the priorities. The chain strengthens the team but does not replace the product owner: the critical decisions are made by a human with the right competence.
When the chain is useful
- The product decision must rest on interviews, not on opinions.
- The requirements conflict, and the team is stuck between compromises.
- Before release the interface needs to be checked for honesty and accessibility.
Questions and answers
Is it required to run all three agents?
No: each agent is useful on its own. The chain is needed when a solution travels the path from data to checking the consequences.
What is passed between the agents?
Structured artifacts: a report with the opportunity map, the contradiction map and the solution directions — each with sources.
Can the data be corrected between the stages?
Yes, at every handoff a human checks and adjusts the materials before the next agent.
Which roles make up the Golden Trio?
The CustDev Agent is the researcher, Super Oracle is the inventor, and Agent-Muse is the auditor.
Why is there a human between the agents?
At every handoff a human checks and adjusts the materials before they reach the next agent.
Can a stage be skipped?
Technically yes, but then the next agent will receive unchecked inputs; control at the handoffs is part of the method.
What do you have in hand after the Trio?
A research report with the opportunity map, a contradiction map with directions and tests, and an audit with the priorities of fixes.
How is the Trio connected to security?
Each agent works with limited permissions and a journal; disputed decisions stop and wait for confirmation.
Where to see the chain in action?
On the Golden Trio scenario page and in a demo run on synthetic data.