Designing Agent Teams

Generate multi-agent coding teams with optimal model assignments. Outputs a reasoning document (AGENTS.md) plus agent instruction files (.agents/) and platform-specific configs where applicable.

Determine the mode

1. create-existing — Analyse an existing codebase, generate a tailored agent team
2. create-new — From a project brief or tech stack, generate a starter agent team
3. add — Add one or more agents to an existing team (e.g., a new dependency or cross-cutting concern)
4. review — Evaluate an existing agent team config, suggest refinements

Creating for an existing codebase? → Follow "Gather inputs: existing codebase" below Starting a new project? → Follow "Gather inputs: new project" below Adding agents to an existing team? → Follow "Gather inputs: add" below Reviewing an existing team? → Follow "Gather inputs: review" below

Gather inputs

Gather what is available from the codebase and the user's request. If the user has not specified priorities, infer them from context (a solo side project likely values speed and cost; a team production app likely values quality). Do not block on missing information — use sensible defaults and note assumptions in AGENTS.md.