arena

# Arena On Codex, read the [platform mapping](../poteto-mode/references/codex-tools.md), including its per-skill notes, before following this skill. On GitHub Copilot, read the [platform mapping](../poteto-mode/references/copilot-tools.md), including its per-skill notes, before following this skill. On Muse, read the [platform mapping](../poteto-mode/references/muse-tools.md), including its per-skill notes, before following this skill. Open a todolist with one entry per phase (Frame, Fan out, Cross-judge, Pick, Graft, Verify) before launching anything. ## Phase A: Frame Every candidate gets the same prompt, so the prompt is the contract. 1. State the artifact each candidate is producing. 2. Derive the rubric, 3-6 concrete gradeable criteria for what success looks like on *this* task. It 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](#models). An `auto` or `inherit-parent` entry in this 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. If the family has no default, or the `Agent` tool rejects that default, use the closest valid slug of the same family from its error message, or drop the seat and say so when there is none. If dropped seats leave fewer than two runners, add runners until there are two, or one when the line has one entry. Run each added runner on a remaining runner's model, or omit `model` so it runs on the parent model when none remain, and say so. Spawn more runners when the arena covers multiple design directions. Use the same model N times when the work is generation-bound rather than judgment-sensitive. Pick each subagent's reasoning level for the job, per the Subagents section of the **poteto-mode** skill. 4. Give each candidate its own output 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 the artifact and a short rationale naming the alternatives it considered and what it rejected. If a candidate produces no output, proceed with N-1 and note the dropout. ## 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](#models). Spawn one readonly judge subagent. The judge is never from the parent's family, as Phase A step 3 defines families. Try only the entries from another family, in pool order. An `auto` or `inherit-parent` entry is the parent model, so it never judges. If the `Agent` tool rejects an entry, use a valid slug of the same model at the closest level or speed from the error message, or move to the next entry when there is none. If no entry from another family runs, try the default pool's entries from another family the same way. If none of those runs, skip the judge, pick the base alone in Phase D, and say so. If you do not know the parent's model, try the pool's real slugs in pool order and say that the family check did not run. The judge sees the rubric and the candidates by path label, scores each criterion, and recommends a base with rationale. It runs in parallel with your reading in Phase D. Don't spawn it while candidates are still writing. ## Phase D: Pick a base Read every candidate end to end before picking. Score each against the rubric criterion by criterion, not on holistic feel, and compare with the cross-judge when Phase C ran one. 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 a future maintainer can extend most easily without breaking invariants. When two feel tied, prefer the cleaner boundary or smaller API, per the Laziness Protocol. ## Phase E: Graft Walk each losing candidate once more for what is worth porting into the base, 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, so the result stays coherent under one mental model. Don't paste mechanically. If the candidates converge on one shape, that is strong agreement. Note it and ship the consensus shape. No graft is needed. If they wildly diverge, Phase A was under-specified. Reframe and re-run rather than averaging the divergence. ## Phase F: Verify Verify the synthesized artifact like any other output, per the **prove-it-works** principle skill. If verification finds a problem the arena missed, either Phase A was wrong (re-frame and re-run) or a candidate caught it and you missed the graft (go back to Phase E). Don't paper over it. ## Outputs One synthesized artifact, with a short synthesis note alongside naming the base and why, the cross-judge's verdict or that Phase C skipped the judge, each graft with its source candidate, what was rejected and why, any convergence or dropouts, 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. On Claude Code, `session` sets no effort. On other runtimes, follow the platform mapping for `session`. 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`. Pass the level as the dispatch's effort field and keep the `subagent_type` you would otherwise use. That is `effort` on the Claude Code `Agent` call and `reasoning_effort` on Codex's `spawn_agent`. On Muse it is `effort` on a Workflow child; a `subagent_spawn` dispatch keeps the session's effort.