adversarial-swarm-analysislisted

# Adversarial Swarm Analysis A template for questions where the answer depends on which lens you look through. One agent asked "should we kill the free tier?" returns a balanced non-answer. A swarm of agents with *conflicting interests*, each forbidden from being balanced, returns a spectrum — and the shape of their disagreement is the actual finding. The workflow is domain-agnostic. Only the roles change between "should we ship this pricing change", "why is this test flaky", "which of these five ideas is worth building", and "what did this experiment actually show". ## When to Use Use it when the question is **open, consequential, and lens-dependent**: - A decision with real tradeoffs (pricing, architecture, hiring, go/no-go). - Idea generation where you want range, not the first obvious answer. - Strategy evaluation (a trading rule, a launch plan, a migration approach). - Troubleshooting where the cause is genuinely unknown and competing hypotheses deserve independent champions. - Reflecting on an experiment or incident where the interpretation is contested. - Any moment you catch yourself writing "it depends on…" — that is the swarm's cue. ## When NOT to Use - **The question has a lookup answer.** "What port does the server bind to?" needs a grep, not a swarm. - **The work is mechanical.** Renames, migrations, test fixes — use `implementer` agents or do it yourself. - **The domain already has a specialist workflow.** For RFCs and design docs, use `rfc-iterative-revi