multi-agent-workflow

You were invoked as a slash command. The user's input:

$ARGUMENTS

Use that input to fill this prompt's variables (take the main content, topic, or task from it; ask only if a required value is missing and not supplied), then follow the prompt exactly.

Design a multi-agent workflow for: {task}

Constraints: {constraints}

Work through the design:

1. Decide if it even needs multiple agents. Multi-agent adds coordination cost; use it when the task genuinely benefits from parallel coverage, independent perspectives, or a scale one context cannot hold. If a single agent with good tools suffices, say so.
2. Decompose the work into roles: what each agent does, its single clear responsibility, and what it needs (tools, context, inputs). Avoid overlap; each agent owns one job.
3. Choose the topology: pipeline (each stage feeds the next), parallel fan-out (independent agents, results merged), orchestrator-workers (a coordinator delegates and synthesizes), or a loop (until a condition). Match it to the task's dependency structure.
4. Define the handoffs: exactly what each agent passes on (structured, not loose prose), so the next agent has what it needs (see agent-handoff-protocol).
5. Add verification: an adversarial or independent checking step for anything that must be correct, because agents are confidently wrong (a reviewer agent, a majority vote, a validation gate).
6. Bound it: cost/latency limits, max iterations, and what happens on failure (fallback, escalation to a human).

Rules: simplest topology that does the job; do not fan out for its own sake. Every agent needs a clear role and a verification story for its output. Name the coordination cost honestly. If the task does not need multiple agents, the right answer is to say that.