arena

Arena

On Codex, read the platform mapping, including its per-skill notes, before following this skill.

On GitHub Copilot, read the platform mapping, including its per-skill notes, before following this skill.

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 pstack-models.md. If the sheet or that line is missing, run one each on the defaults in Models. An auto or inherit-parent entry in this line or the cross-judge line means the parent model, so omit model for it. If the Agent tool rejects a configured entry, run that seat on its family's default and say so. Families go by model name, such as Opus, Fable, or Sonnet. With no family match, use the single-role default in Models. If it rejects a default, use the closest valid slug of the same family from its error message. Spawn more when the arena covers multiple design directions. Same model N times when the work is generation-bound rather than judgment-sensitive.
4. Assign output paths. Each candidate writes to its own location (a git worktree 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 run_in_background: true, each with the task, the path to the shared grounding, its own output path, and instructions to produce both the artifact and a short rationale.

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 pstack-models.md. If the sheet or that line is missing, choose from the runner defaults in Models. Prefer a different model family from the parent's. Spawn one readonly judge subagent on that model. 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.

Models

Role defaults, stamped from plugins/pstack/models.json (edit there, rerun tools/generate.mjs). A matching role line in the pstack-models.md override sheet overrides each at runtime; /setup-pstack writes it and lists its path per runtime.

• arena runners: opus, fable, sonnet, haiku
• arena cross-judge pool: opus, fable, sonnet, haiku

Reasoning effort

A role value in the override sheet may name a reasoning effort after its model, as in opus @xhigh. Levels on Claude Code: low, medium, high, xhigh, max. Which ones apply depends on the model. A value without @ takes the sheet's default effort line, a level or session, and session when the sheet has no such line. session sets no effort, so the dispatch is the usual one. Strip the suffix before reading the model: inherit-parent or auto still omits model at every level, and a model name is passed as model. On Claude Code, a level picks the effort agent from the subagent_type you would otherwise use. pstack:poteto-agent becomes subagent_type: "pstack:poteto-agent-<level>". general-purpose, or no subagent_type, becomes subagent_type: "pstack:effort-<level>". The effort agents set only effort, so the model you pass still decides the model. On Codex, pass the level as spawn_agent's reasoning_effort and keep the usual instructions.