Official agent skill

Arena Parallel Candidates

by cursor in cursor/plugins

Runs several parallel attempts at the same task, picks the strongest as a base, grafts the best ideas from the others into it, then verifies the merged result.

OfficialNo licenceAuto-check passedAgent Workflows

Install Arena Parallel Candidates

skills CLI
$ npx skills add cursor/plugins --skill arena -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install cursor/plugins arena --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/cursor/plugins.git skills-src && mkdir -p .claude/skills && cp -r skills-src/pstack/skills/arena .claude/skills/arena && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
arena
GitHub stars
10k
Used in
8 other repos
Token cost
~1.2k tokens
SKILL.md length
727 words
Files
1
Skills in repo
99
Repo updated
First seen
Licence
None found

At a glance

Runs several parallel attempts at the same task, picks the strongest as a base, grafts the best ideas from the others into it, then verifies the merged result.

  • Works in 6 steps: Frame → Fan out → Cross-judge → …
  • A non-trivial artifact where a single attempt may fix the wrong design
  • SKILL.md covers Start, Phase A: Frame, Phase B: Fan out and Phase C: Cross-judge, plus 4 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Arena is for work where one attempt could lock in the wrong shape. The agent opens a todo list with six phases: frame, fan out, cross-judge, pick, graft and verify. Framing fixes the shared prompt, derives a rubric of three to six gradeable criteria that only the picker sees, chooses which models run the candidates from a rules file or defaults, and gives every candidate its own output location, a git worktree where possible.

In the fan-out phase all subagents start in one message in the background. Each gets the task, a path to shared grounding and its output path, and returns the artifact plus a short rationale naming the alternatives it considered and rejected. If one candidate produces nothing, the run continues with the rest and notes the dropout. The agent reads every candidate end to end before choosing a base. The excerpt cuts off before the later phases.

When your agent uses it

  • A non-trivial artifact where a single attempt may fix the wrong design
  • Comparing several design directions and combining their best parts
  • You ask to arena a task or throw it in the arena

Example prompts

  • “Arena this: draft the public API for the new caching layer.”
  • “Throw the onboarding email sequence in the arena and merge the strongest parts.”
  • “Run parallel candidates for the settings page redesign and pick a base to build on.”

Requirements

  • An agent that can start background subagents

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

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

What it can do on your machine

Read from SKILL.md and the folder at commit 9f451cf. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Arena Parallel Candidates loads about 1.2k tokens when it runs. Until then it costs about 61 tokens; SKILL.md has 727 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~61
When it runs · the whole SKILL.md, loaded when a task matches
~1.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

Without a licence we can't republish the file, so here is its outline and opening line. It has 727 words (~1,235 tokens).

“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.”

— opening of SKILL.md by cursor
name
arena
disable-model-invocation
true

Read the full SKILL.md on GitHub

Files

Just SKILL.md in pstack/skills/arena of cursor/plugins.

Open the folder on GitHubat commit 9f451cf

Used in 8 other repositories

We found 8 copies of this SKILL.md (exact, near-identical or edited) in other folders, from 8 other GitHub owners. This page covers the copy in cursor/plugins, which our catalogue first saw on October 7, 2026.

Compare with similar skills

Arena Parallel Candidates next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Arena Parallel Candidates compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Arena Parallel Candidates this skillcursor/plugins10k8 repos~1.2kAutomated safety check: PassNone
Paseo Advisor Second Opiniongetpaseo/paseo20k1 repos~756Automated safety check: PassCustom licence
O2 Review Loopopenobserve/openobserve22k—~3.7kAutomated safety check: PassAGPL-3.0
Harness Agent Team Designerrevfactory/harness9.1k—~4.5kAutomated safety check: PassApache-2.0
ClawTeam Multi-Agent Swarmwin4r/ClawTeam-OpenClaw1.5k1 repos~2.9kAutomated safety check: PassMIT
Sub-Agent Delegationcodewhale-hq/Codewhale41k—~790Automated safety check: PassMIT

Similar skills

  • Launches one separate agent through Paseo to give a second opinion on the current task, with a self-contained briefing and no permission to edit files.

    20k GitHub starsUsed in 1 repo~756 tokens
    Agent WorkflowsAuto-check passed
  • O2 Review Loop

    openobserve/openobserve

    Splits a change into planner, coder and independent reviewer roles: you confirm a spec, a subagent implements it, and a separate reviewer checks each round's local WIP commit.

    22k GitHub stars~3.7k tokensUpdated today
    Agent WorkflowsAuto-check passed
  • Harness Agent Team Designer

    revfactory/harness

    Designs a project-specific agent harness: defines specialist agents, writes the skills they follow, picks an execution mode and model for each, and keeps the setup maintained.

    9.1k GitHub stars~4.5k tokensUpdated 9 days ago
    Agent WorkflowsAuto-check passed
  • ClawTeam Multi-Agent Swarm

    win4r/ClawTeam-OpenClaw

    Launches a swarm of specialist Hermes agents in git-worktree-isolated tmux windows with a kanban board and file-based inboxes, using built-in templates like hedge-fund and code-review.

    1.5k GitHub starsUsed in 1 repo~2.9k tokens
    Agent WorkflowsAuto-check passed
  • Sub-Agent Delegation

    codewhale-hq/Codewhale

    Guides when and how to split multi-step coding, research or verification work into focused sub-agent runs while the parent keeps integration and final checks.

    41k GitHub stars~790 tokensUpdated today
    Agent WorkflowsAuto-check passed
  • Parallel Agents

    xenitV1/Antigravity-Workflows

    Native multi-agent orchestration using Claude Code's Agent Tool.

    130 GitHub starsUsed in 9 repos~1.3k tokens
    Agent WorkflowsAuto-check passed

More from cursor/plugins

All 99 skills in this repo
  • Official

    Keeps a TSV decision log for long or unattended agent runs, one row per decision with what, why, evidence and result, so a reviewer can check the work later.

    10k GitHub starsUsed in 9 repos~1.6k tokens
    Auto-check passed
  • Official

    Digs into why code is shaped the way it is by checking git history, pull requests and connected tools in parallel, then reporting a cited read on the tradeoffs.

    10k GitHub starsUsed in 9 repos~2.6k tokens
    Auto-check passed
  • Official

    Starts three parallel reviewer subagents over the current conversation transcript, then turns their findings into concrete edits to existing skills.

    10k GitHub starsUsed in 5 repos~1.2k tokens
    Auto-check passed
  • Official

    Applies four layers of technical-writing rules to docs, RFCs, readmes, PR descriptions and commit messages so a tired engineer follows them on the first read.

    10k GitHub starsUsed in 10 repos~2.4k tokens
    Auto-check passed
  • Advisor Mode

    cursor/plugins

    Official

    Adds a second, stronger model that the main agent consults before major decisions, when stuck and before finishing, controlled by /advisor commands.

    10k GitHub stars~2.6k tokensUpdated today
    Auto-check: notes
  • Official

    Prepare PRs for review by cleaning noisy history, improving PR descriptions, and adding reviewer guidance without changing code behavior.

    10k GitHub starsUsed in 3 repos~569 tokens
    Auto-check passed

Categories

Questions about Arena Parallel Candidates

What does Arena Parallel Candidates do?

Runs several parallel attempts at the same task, picks the strongest as a base, grafts the best ideas from the others into it, then verifies the merged result. Arena is for work where one attempt could lock in the wrong shape. The agent opens a todo list with six phases: frame, fan out, cross-judge, pick, graft and verify.

When should I use Arena Parallel Candidates?

Arena Parallel Candidates fits situations like: A non-trivial artifact where a single attempt may fix the wrong design; comparing several design directions and combining their best parts; you ask to arena a task or throw it in the arena.

How do I install Arena Parallel Candidates in Claude Code?

Run `npx skills add cursor/plugins --skill arena -a claude-code`. Or copy the skill folder (pstack/skills/arena in cursor/plugins) into .claude/skills/arena in your project. Claude Code loads it when a task matches its description.

How do I install Arena Parallel Candidates in Codex?

Run `npx skills add cursor/plugins --skill arena -a codex`. Or copy the skill folder (pstack/skills/arena in cursor/plugins) into .agents/skills/arena in your project. Codex loads it when a task matches its description.

Can I use Arena Parallel Candidates in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add cursor/plugins --skill arena -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/arena, .gemini/skills/arena, .github/skills/arena and .opencode/skills/arena in your project.

What does Arena Parallel Candidates need to run?

SKILL.md names no scripts, command-line tools or credentials: Arena Parallel Candidates is instructions for the agent only. Our summary lists: An agent that can start background subagents.

Does Arena Parallel Candidates access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Arena Parallel Candidates safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Arena Parallel Candidates use?

No licence was found for Arena Parallel Candidates or its repository. Without one, default copyright applies: ask the author before reusing or redistributing it.

How many tokens does Arena Parallel Candidates use?

About 1.2k tokens (SKILL.md is roughly 4.9k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Arena Parallel Candidates?

Skills that share tags, products or a category with Arena Parallel Candidates: Paseo Advisor Second Opinion (getpaseo/paseo, 20k stars), O2 Review Loop (openobserve/openobserve, 22k stars), Harness Agent Team Designer (revfactory/harness, 9.1k stars) and ClawTeam Multi-Agent Swarm (win4r/ClawTeam-OpenClaw, 1.5k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Arena Parallel Candidates?

cursor (a GitHub organization, an official publisher) maintains it in cursor/plugins, which has 10,130 GitHub stars. The repository holds 99 skills in this directory. The repository was last updated on October 6, 2026.

Source: cursor/plugins on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.