test-team-coordination
About
This skill executes test scenarios against AI teams to validate and compare their coordination patterns. It observes team behaviors, evaluates acceptance criteria, and automatically generates structured RESULT.md reports. Use it for validating team performance, comparing coordination approaches, or establishing performance baselines during development.
Quick Install
Claude Code
Recommendednpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/test-team-coordinationCopy and paste this command in Claude Code to install this skill
Documentation
測隊協
行測場於目隊。察協模為、評納則、計則、生 RESULT.md 於 tests/results/。
用
- 驗隊協模生期為→用
- 改隊定或代後行構測→用
- 比協模、同場異隊→用
- 為隊組立基效標→用
- 加新代或變員後回歸測→用
入
- 必:測場檔路(如
tests/scenarios/teams/test-opaque-team-cartographers-audit.md) - 可:行 ID 蓋(默:
YYYY-MM-DD-<target>-NNN自生) - 可:隊大蓋(默:自場前題)
- 可:略範變(默:否——若定則注範變)
行
一:載驗測場
1.1 讀入指之測場檔。
1.2 析 YAML 前題取:
target— 待測之隊coordination-pattern— 期模team-size— 員生數- 納則表
- 計則(若有)
- 真值資(若有)
1.3 驗場檔諸需段:
- 標
- 前況
- 任(含主任子段)
- 期為
- 納則
- 察議
得:場檔載、析、含諸需段。
敗:檔失或不可析→止以誤訊識失檔或誤段。可選段(則、真值、變)缺→記缺而續。
二:驗前況
2.1 走場諸前況格。
2.2 檔存察用 Glob 驗。
2.3 登數察析 _registry.yml 比 total_* 於碟實檔數。
2.4 枝/git 態察行 git status --porcelain 與 git branch --show-current。
得:諸前況滿。
敗:前況敗→記為 BLOCKED 於果。決續(軟前況)或止(硬前況如失目隊檔)。文決。
三:載協模則
3.1 讀 tests/_registry.yml 覓 coordination_patterns 配場 coordination-pattern 之值。
3.2 取此模 key_behaviors 列。
3.3 此為察清——各於行中察、記察/不察。
得:模要為載、備察。
敗:協模未於登定→用場期為段為唯察源。記警。
四:行任
4.1 建果目:tests/results/YYYY-MM-DD-<target>-NNN/。
4.2 記 T0(任始時)。
4.3 自 teams/<target>.md 讀目隊定、取 CONFIG 塊、活隊:呼 TeamCreate 含隊名、生伴用各員之 subagent_type、自 CONFIG tasks 列建任。用場之隊大。傳場任段之主任提原文。
4.4 察隊執相。記時於:
- T1:形察/任分畢
- T2:角分顯
4.5 場定範變觸 而 skip-scope-change 為 false:
- 待至相二(角分)顯
- 記 T3(範變注時)
- 經 SendMessage 送範變提於隊
- 記 T4(範變吸——角調顯)
4.6 續察至隊交出。
- 記 T5(整始)
- 記 T6(末報交)
4.7 捕隊全出。
得:隊行任過協模相。諸轉時記。範變(若可)注吸。
敗:隊不生出→記敗點與誤訊。隊滯→記末察相與超時。以部果進評。
五:評模為
5.1 各步三要為定行中是否察:
- 察:隊出或協明證
- 部:某證而不全或歧
- 不察:無證
5.2 各場期為段任特為施同評。
5.3 記發於察日。
得:諸或多模特與任特為察。
敗:未察為發、非測程之敗。準記——示協模未全顯。
六:評納則
6.1 走場諸納則。
6.2 各則定:
- PASS:明滿含可察證
- PARTIAL:部滿(計閾以 0.5 重)
- FAIL:有機而未滿
- BLOCKED:不可評(前況敗、隊超時等)
6.3 場含真值資→驗報發於彼:
- 各類算準率
- 標誤正與誤負
6.4 場含計則→各維 1-5 計含簡證。
6.5 算總度:
- 納:X/N 過(PARTIAL 計 0.5)
- 閾:>= 場定閾→PASS
- 則總:X/Y 點(若可)
得:諸納則有定。總度算。
敗:少於半則可評(BLOCKED 過多)→測行不確。文何故並薦修前況後重行。
七:生 RESULT.md
7.1 用場察議之記模建 tests/results/YYYY-MM-DD-<target>-NNN/RESULT.md。
7.2 填諸段:
- 行元(察、時、長)
- 相日含諸記時
- 角生日(為適/隊測)
- 納則果表
- 則計表(若可)
- 真值驗表(若可)
- 要察(敘)
- 教
7.3 隊原出含為附或別檔(team-output.md)於同果目。
7.4 加首總判:
**Verdict**: PASS | FAIL | INCONCLUSIVE
**Score**: X/N criteria (Y/Z rubric points)
**Duration**: Xm
得:全 RESULT.md 含諸段填與明判。
敗:果檔不可書→出果至 stdout 為退。評資永不當失。
驗
- 測場檔載而諸需段在
- 前況驗(或文為 BLOCKED)
- 協模要為自登載
- 隊生而任交
- 範變注於正時(若可)
- 諸模特為評(察/部/不察)
- 諸納則有定(PASS/PARTIAL/FAIL/BLOCKED)
- 真值驗畢(若可)
- RESULT.md 生含諸段填
- 總判算記
忌
- 評出質非協:此技測隊何協、非任出全否。協善而唯覓 7/9 斷參之隊仍顯模
- 範變注過早:待至角分明顯乃注範變。過早→隊未分、無可適者
- 混員出於隊出:不透隊當示合一出。見個員報→為不透之發、非測基患
- 真值精配:真值數為近。評發於正範、非配精
- 忘記時:時要為計相長與適速。事生時設、勿後
參
review-codebase— 深碼覆補隊層測review-skill-format— 驗個技式(此技驗隊協)create-team— 建此技測之隊定evolve-team— 依測發演隊定test-a2a-interop— A2A 協合規類測模assess-form— 不透隊長內用之態變評
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
polymarket
MetaThis skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.
creating-opencode-plugins
MetaThis skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
