01 / GETTING STARTED
Quick start
Start with a specific decision rather than a broad topic. Add the context a thoughtful reviewer would need, and say what kind of outcome would be useful.
- Frame the decision. State the question, objective, constraints, and timeline.
- Choose perspectives. Include viewpoints that differ in incentives, expertise, or risk tolerance.
- Review the debate. Examine disagreements, assumptions, and claims that need evidence.
- Capture next steps. Summarize options, open questions, and follow-up research.
02 / WORKFLOW
Workflow structure
A useful review keeps these elements distinct. The specific implementation can vary, but the reasoning should remain legible.
The decision or question to examine.
Different roles or positions that challenge the initial framing.
Risks, alternatives, assumptions, and uncertainties found during the debate.
A concise summary to support human review and follow-up.
03 / EXAMPLE
Example configuration
This is an illustrative configuration showing the information you might collect for a scenario review. It is not a live API contract.
{
"question": "Should we enter a new market?",
"context": [
"limited launch budget",
"uncertain customer demand",
"six-month review horizon"
],
"perspectives": [
"customer advocate",
"skeptical analyst",
"operations lead"
],
"review": [
"assumptions",
"risks",
"alternatives",
"evidence gaps"
]
}04 / GOOD PRACTICE
Get more useful reviews
- Be specific. Narrow the question to the decision you actually need to make.
- Invite real disagreement. Perspectives should differ meaningfully, not just restate the same point.
- Ask for uncertainty. Request assumptions, confidence limits, and what evidence might change the conclusion.
- Verify important claims. Treat outputs as analytical support, not as a substitute for research or expert review.
05 / QUESTIONS
FAQ
Does ToxicAgents make the final decision?
No. The goal is to organize competing viewpoints and surface trade-offs. The human decision-maker remains responsible for evaluating the findings.
Should I trust every generated claim?
No. Check consequential claims against reliable sources, especially in legal, financial, medical, safety, and other high-stakes contexts.
Can I use this workflow with my team?
The workflow can be adapted for team reviews. Contact us to discuss the intended process, collaboration needs, and technical requirements.