Arena

---
name: arena
description: "Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it. Use for /arena, 'arena this', 'throw it in the arena', or when one attempt at a non-trivial artifact would lock in the wrong shape."
disable-model-invocation: true
---

# Arena

Fan out N parallel attempts at the same task. Read every candidate end to end. Pick the strongest as the base. Graft the best ideas from the others into it. Verify the synthesized result.

## Start

Open a todolist with one entry per phase before launching anything.

1. Frame
2. Fan out
3. Cross-judge
4. Pick
5. Graft
6. Verify

## Phase A: Frame

The N candidates will receive the same prompt, so the prompt is the contract.

1. State the artifact each candidate is producing.
2. Derive the rubric. State what success looks like for *this* task, then turn it into 3-6 concrete gradeable criteria. The rubric is the picker's tool in Phase D. Candidates only see the task.
3. Pick the runners. Use the `arena runners` line in `~/.claude/pstack-models.md` (a `.claude/pstack-models.md` in the repo root wins over it). Read the file if it exists. If the file or that line is missing, default to one each on `fable` and `opus` (runners write code, so no `sonnet` seat). An `inherit` entry (or the older `auto` / `inherit-parent`) in this line or the cross-judge line means the parent model: omit `model` on a `general-purpose` seat, and pass the parent's own model name on a `poteto-agent` seat, since omitting it there falls back to the agent file's `opus`. The Agent tool's `model` accepts only `fable`, `opus`, `sonnet`, and `haiku`. Any other entry (a Cursor slug, a full model ID) runs that seat on `opus`, and say so. Spawn more when the arena covers multiple design directions. Same model N times when the work is generation-bound rather than judgment-sensitive.

   Every default seat is a Claude model; the two models diverge in what they produce. Optional extra seat: if `codex` is on PATH, add one candidate as a `codex exec` run via Bash (Bash `run_in_background: true`), working in its own output path (a `git worktree add` checkout when it writes into the repo) like any other candidate. Never require it.
4. Assign output paths. Each candidate writes to its own location (a git worktree via `isolation: "worktree"` on its Agent call where possible, otherwise `/tmp/arena-<slug>/candidate-<n>/`), per the **separate-before-serializing-shared-state** principle skill.

## Phase B: Fan out

Spawn all N subagents in one message with the Agent tool, each on its runner's `model`. A candidate that writes code uses `subagent_type: "poteto-agent"` (its agent file fixes effort at `high`) with `isolation: "worktree"`. Other candidates use `general-purpose`. Give each the task, the path to the shared grounding, its own output path, and instructions to produce both the artifact and a short rationale. A worktree candidate ends its report with its worktree path and branch. Those become the path labels for Phases C and D.

Each rationale names the alternatives the candidate considered and what it rejected.

If a candidate fails to produce output, proceed with N-1 and note the dropout in the synthesis record.

## Phase C: Cross-judge

After all Phase B candidates complete, choose one model from the `arena cross-judge pool` line in the same models file. If the file or that line is missing, choose from `fable`, `opus`, `sonnet`. Prefer a different model from the parent's. An `inherit` entry counts as the parent's model, so skip it when the pool has another entry. Spawn one `general-purpose` judge subagent on that model and tell it not to edit any file. It sees the rubric and the candidates by path label, scores each criterion, and recommends a base with rationale. It runs in parallel with the parent's reading in Phase D, not with the candidates themselves. Don't spawn the judge while candidates are still writing.

## Phase D: Pick a base

Read every candidate end to end before picking.

Score each candidate against the rubric criterion by criterion, not on holistic feel. Compare against the cross-judge. Agreement on the base confirms the pick. Disagreement means one of you is biased or the rubric was ambiguous. Read both rationales before deciding.

Pick the base on which candidate a future maintainer can extend most easily without breaking invariants. Prefer the cleaner boundary or smaller API when two feel tied, per the Laziness Protocol.

Record the pick and the reason in a short synthesis note alongside the base artifact, including the cross-judge's verdict.

## Phase E: Graft

Walk each losing candidate once more and identify what is worth porting into the base. The signal is usually one or two things per candidate, not most of it.

Fold each graft in by hand, per the **redesign-from-first-principles** principle skill. Don't paste mechanically. The result has to remain coherent under one mental model.

Record what was grafted, from which candidate, and what was rejected and why.

When N candidates converge on the same shape, that is a strong agreement signal. Note the convergence in the record and ship the consensus shape. No graft is needed. When N candidates wildly diverge, Phase A was under-specified. Reframe and re-run rather than averaging the divergence.

## Phase F: Verify

The synthesized artifact has to hold up under the same scrutiny as any other output, per the **prove-it-works** principle skill.

If verification surfaces a problem the arena did not catch, either Phase A was wrong (re-frame and re-run) or one candidate caught it and you missed the graft (go back to Phase E). Don't paper over.

## Outputs

One synthesized artifact. One short synthesis note alongside, naming the base, the grafts (with source candidate), the rejections, the dropouts if any, and the verification result.