Agent

Red-team challenge loop

Runs a structured adversarial challenge against the draft argument — generating the strongest plausible counter-arguments and stress-testing each recommendation.

In / outArgument structure + key recommendations → adversarial challenge log (counter-arguments, logic gaps, mitigation options per challenge)

You might say…

I want to hear the hard questions before the client does, not in the room — the red-team loop is like hiring a sceptic for an hour.

What it does

Autonomously attacks the recommendation set from a sceptical-client stance across multiple rounds — probing logical gaps, missing data, and shaky assumptions, escalating until no new challenge lands. Used to pressure-test the full argument before presentation; orchestrates iterative challenge rather than a single judgement.

Trigger: Use after the narrative arc is stable and before the final deliverable is built; provide the argument structure and key recommendations.

Autonomy: Acts with approval — does the work, checks in on the exceptions.

Recognise the problem?

The primitives are the commodity part. The fastest next step is a conversation about composing them into something that works for you.

Start a conversation