arena
Use for Pstack design or artifact bakeoffs, blind judging, base selection, and grafting useful alternatives.
How do I install this agent skill?
npx skills add https://github.com/christofferbergj/agent-skills --skill arenaIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill implements a multi-agent 'arena' workflow for design comparison and synthesis. While it is functionally well-structured, it carries a low-severity risk of indirect prompt injection because it instructs the agent to read and evaluate generated subagent outputs without using boundary markers or sanitization.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Arena
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.
- 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 from /swarm's
arena-runnerspanel, by default Fable athigh, Opus athigh, and the other vendor's model, seated as its execution contract says: Fable and Opus as native subagents in Claude Code and the third seat through the counterpart route, or the third seat on the session's model in Codex and Fable and Opus through the Claude counterpart. Anautoorinherit-parententry is the parent model. If a seat cannot run or the host rejects its model, follow /swarm's rejected-model rule or drop the seat, keep at least two runners, and say so. Spawn more for multiple design directions, and repeat one model when the work is generation-bound, not judgment-sensitive. Pick each subagent's reasoning level through /swarm. - Give each candidate its own output location (a git worktree where possible, otherwise
/tmp/arena-<slug>/candidate-<n>/), per /principle-separate-before-serializing-shared-state.
Phase B: Fan out
Spawn all N subagents at once, in the background, through /swarm, 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 one readonly judge subagent on the counterpart through /swarm's cross-vendor critic route. Upstream's judge pool is Opus, then the other vendor's model, and the judge is never from the parent's vendor, with Fable counting as Claude, so the judge is Claude Opus under Codex and Codex under Claude Code. Pass it high effort, the level of upstream's pool, moved by the budget. Take only the counterpart routes, T3 and then the CLI, never the same-vendor critic. If the counterpart resolves to fallback or neither route runs, skip the judge, pick the base alone in Phase D, and say so. While you read in Phase D, the judge scores each candidate, by path label, on each rubric criterion and recommends a base.
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 /principle-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 /principle-redesign-from-first-principles, 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 /principle-prove-it-works. 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.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/christofferbergj/agent-skills/arena">View arena on skillZs</a>