跳到正文
cccskills
免费在 GitHub 公开

multi-agent-tournament-scorecard

Use when agent strategies need a reproducible finite tournament under a frozen evaluation protocol. Not for open-ended or infinite matchups: use a continuous benchmark.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md3.4 KB
  • agents/openai.yaml185 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Multi-agent tournament scorecard

Contract

FieldBound contract
TriggerAgent strategies need a reproducible finite tournament under a frozen evaluation protocol.
AuthorityReversible local writes only. Run records and receipts under .outline/loops/; rollback is deletion. No VCS, credential, paid, published, deployed, or remote mutation.
Side effectA reproducible tournament scorecard persisted as a run record.
DoneThe run ends at exactly one declared terminal class with a persisted run record and receipt.json: on success, the complete matrix with move records, scores, and hidden-identity validation; on stalled, the named matchup and stopping reason; on blocked, the named violation or mismatch and no published scorecard; on exhausted, the completed matchups and the remainder, with no success claim.

Inputs

  • Matchup matrix (required): the explicit finite set of agent pairings and rounds, frozen before the tournament begins.
  • Evaluation protocol (required): the frozen rules that govern each matchup: move validation, scoring, and identity disclosure.
  • Agent strategies (required): the strategies under test, each identified and version-pinned.

Procedure

  1. Bind the declared matchup matrix and freeze it before any run. Done when: the matrix is recorded in writing and no matchup has been executed yet.
  2. Execute the tournament inside the bound: run each matchup in the frozen matrix, record every move, and compute scores per the evaluation protocol. Done when: every matchup in the matrix has a recorded outcome or a non-success terminal applies.
  3. Validate hidden-identity assignments: confirm each assignment is consistent across the matrix and no identity leaked before disclosure. Done when: every identity assignment is verified or the mismatch is named.
  4. Select the terminal class, exactly one: success only when the matrix is complete and step 3 verified every identity; otherwise stalled for an incomplete matchup, blocked for an invalid run or a hidden-identity mismatch, exhausted for budget exhaustion before the matrix completes. Budget exhaustion is never success. Done when: exactly one terminal class is selected and recorded.
  5. Persist the result: write the run record to .outline/loops/<slug>/<run_id>/ when durable, and emit receipt.json before returning. Done when: the receipt file exists and contains every required field.

Failure and recovery

  • Incomplete matchup: a matchup in the matrix did not finish. Terminal stalled. Name the matchup and the stopping reason.
  • Invalid run: the evaluation protocol was violated during a matchup. Terminal blocked. Name the violation and the affected matchup.
  • Budget exhausted before matrix completes: terminal exhausted. Report the matchups completed and the remainder. Do not claim success.
  • Hidden-identity mismatch: an identity assignment is inconsistent or leaked. Terminal blocked. Name the mismatch; do not publish the scorecard.

Output

A tournament scorecard: terminal class (success, stalled, blocked, exhausted), completed matchups, move records, scores, identity validation, and the receipt path, ordered by the procedure steps that produced them.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Use when explaining System V AMD64, ARM AAPCS, RISC-V psABI, stack frames, variadic calls, or FFI register rules. Not for the Rust FFI binding layer: use rust-ffi.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin372026年9月29日 更新

Use when configuring ADC sampling time, DMA-driven ADC, calibration, or DAC channel setup on bare-metal MCUs. Not for the DMA stream itself: use dma-baremetal.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin372026年9月29日 更新

af-xdp

免费

Use when creating AF_XDP sockets, configuring UMEM and XSK rings, writing an XDP redirect program, or choosing copy versus zero-copy mode. Not for full kernel bypass: use dpdk.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin372026年9月29日 更新

Use when a completed session needs an agent-environment retrospective. Not for an engineering retrospective from telemetry: use engineering-retrospective.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin372026年9月29日 更新

Use when a redacted, trimmed agent transcript must be appended to a GitHub PR or issue body, with human approval and preview. Not for automated or model-initiated insertion.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin372026年9月29日 更新

agents-md

免费

Use when a repo needs agent setup, AGENTS.md added or made lean, CLAUDE.md audited, or agent instructions scored or pruned. Not for remote, credential, publish, deploy, or irreversible changes.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin372026年9月29日 更新

查看 OutlineDriven 的全部技能

报告此技能的问题