arena
Spawn N parallel candidates at the same task, pick a base, and graft the strongest parts of the losers into it. Use for arena this or competing attempts.
- License 许可
- MIT
- 最近更新
- 2026-10-21
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- arena
- description
- Spawn N parallel candidates at the same task, pick a base, and graft the strongest parts of the losers into it. Use for arena this or competing attempts.
03 SKILL.md 正文
Arena
Open the plan at the top of your reply, one entry per phase (Frame, Fan out, Cross-judge, Pick, Graft, Verify), and update it as each step lands. v2 has no todo tool.
Phase A: Frame
Every candidate gets the same prompt, so the prompt is the contract.
- State the artifact each candidate is producing.
- 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.
- Pick the runners. Spawn the agents for the
arena runnersrole, one subagent per agent ID, fromreferences/roles.mdin thesetup-pstackskill. Two seats by default. Do not pass a model. Each role agent carries its own; the operator sets them with/pstack/setup-pstack. Spawn more runners when the arena covers multiple design directions. Reuse one seat N times when the work is generation-bound rather than judgment-sensitive. - 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 on the arena runners role agent, with 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, spawn the agents for the arena cross-judge pool role, one subagent per agent ID, from references/roles.md in the setup-pstack skill. Two read-only seats. Do not pass a model. Each judge sees the rubric and the candidates by path label, scores each criterion, and recommends a base with rationale. They run in parallel with your reading in Phase D. Don't spawn them 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-judges. Judges agreeing 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-judges' verdicts, each graft with its source candidate, what was rejected and why, any convergence or dropouts, and the verification result.
04 同仓库更多 Skills
why
Trace why code works this way or why a design was chosen, from source control, trackers, docs, and telemetry. Use for rationale, regressions, or postmortems.
unslop
Cut AI tells from any writing.
typescript-best-practices
TypeScript best practices. Use when reading or editing any .ts or .tsx file.
technical-writing
Write or review docs, RFCs, readmes, PR descriptions, and commit messages in Diataxis structure with Google style, STE, and Global English.
teach
Explain a body of work so a person actually understands it, running how and why and weaving both into one clear explanation. Use for teach me this.
tdd
Use only when the user asks for TDD, a failing test, or a regression test, or when a bug has a cheap local test target; skip unclear or costly test paths.