Agent Swarmdocs
GitHub

11 -Swarm gallery

Customer panel

Three stand-in customers - a first-timer, a daily operator, and a skeptical buyer - react to your product's first-run value, friction, and adoption risk in parallel. The orchestrator folds their takes into one call: Fix now, Defer, or Reject.

The roster

The preset runs three bundled agents at once. Each holds a distinct customer perspective so the panel covers the full adoption arc - discovery, repeated use, and long-term commitment.

  • first-time-user - Evaluates onboarding clarity, setup friction, and the moment early value becomes obvious, from the perspective of someone encountering the product for the first time without insider context.
  • busy-operator - Pushes on workflow fit, defaults, copy/paste burden, and time-to-result, from the perspective of a technical operator who needs the tool to save time immediately.
  • skeptical-buyer - Probes adoption barriers, unclear value, weak differentiation, and missing proof, from the perspective of a technical buyer who needs a clear reason to try and keep using the tool.

Goal & decision

The preset's goal is fixed in the bundle:

Identify the clearest customer-facing value, friction, and adoption risk.

Resolution mode is orchestrator: after each round the bundled orchestrator agent reads all three responses and builds the directive for the next round, then synthesizes the final call in the preset's decision vocabulary:

Fix nowDeferReject

Fix now means a blocking friction or missing value signal that should be addressed before shipping. Defer means the issue is real but not blocking. Reject means the proposal does not earn adoption in its current form.

Note

The decision vocabulary comes straight from the preset's decision field and appears verbatim in synthesis.md. You can override it per run with --decision, but the bundled preset ships with Fix now / Defer / Reject.

Run it

Pass the number of rounds as the first positional argument (1-3), then your question. No config needed - the preset wires up the roster and decision vocabulary:

agent-swarm run 2 "Does the onboarding flow make the first successful run obvious enough to earn a second?" \
  --preset customer-panel

All three agents run in parallel. A two-round run lets the orchestrator refocus the panel on contested points before the final synthesis. For a quicker read, one round is often enough:

agent-swarm run 1 "Does the onboarding flow make the first successful run obvious enough to earn a second?" \
  --preset customer-panel
Note

Real harnesses can take longer than the default 120 s dispatch timeout. Pass --timeout-ms 300000 for deeper runs. Use --quiet for one-line-per-event output in CI.

What the synthesis looks like

When the run finishes you'll find a self-contained directory under .agent-swarm/runs/<timestamp>-<slug>/. The synthesis folds all three agents into one call:

Synthesiscustomer-panel · round 2 · 3 agents

Fix now - onboarding hides the one action that shows value.

consensus2 of 3 agree
confidencehigh
Still openDoes the empty state need a sample project, or just a clearer first step?

Confidence is reported as a level - low, medium, or high - not a fabricated percentage. The verdict word comes straight from the preset's decision vocabulary. "Still open" surfaces questions that at least two agents deferred across all rounds.

Where next