Back to Skills

unleash-the-agents

pjt222
Updated 2 days ago
10 views
17
2
17
View on GitHub
Metaai

About

This skill launches multiple AI agents in parallel to generate diverse hypotheses for complex, cross-domain problems where the solution path is unclear. It's ideal when single-agent approaches fail or when broad exploration is needed over deep specialization. The output is a ranked set of hypotheses, refined through convergence analysis and adversarial critique.

Quick Install

Claude Code

Recommended
Primary
npx skills add pjt222/agent-almanac -a claude-code
Plugin CommandAlternative
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternative
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/unleash-the-agents

Copy and paste this command in Claude Code to install this skill

Documentation

縱諸臣

於正領未明之疾,並波而召諸臣以生開放之假設。各臣以其域之鏡而思——卡巴拉者以數值術尋形,武術者擬條件之分,靜觀者居數而識其結。獨立之見之合,乃假設有益之主信。

用時

  • 跨域之疾,正法未明
  • 單臣或單域已滯而無信
  • 疾賴真正多元之見(非徒增算)
  • 需生假設,非執行(執行用團)
  • 重要之決,遺非顯之角者其代真重

  • 必要:問題之要——疾之清述、五以上之具例、何為解
  • 必要:驗之法——如何試假設正否(程驗、家評、或空模較)
  • 可選:臣之子集——納或排之臣(默:諸已錄之臣)
  • 可選:波之大小——每波之臣數(默:10)
  • 可選:出之式——應之結構模(默:假設+理+信+可試之測)

第一步:備其要

書一要而諸臣無論何域皆能解之。含:

  1. 問題之述:所求者何(一二句)
  2. :五以上具入/出之例或數點(多益佳——三於諸臣為少而難尋形)
  3. 已知之限:所已知者、所已試者
  4. 成之準:如何識正之假設
  5. 出之模:應之精式
## Brief: [Problem Title]

**Problem**: [1-2 sentence statement]

**Examples**:
1. [Input] → [Output]  (explain what's known)
2. [Input] → [Output]
3. [Input] → [Output]
4. [Input] → [Output]
5. [Input] → [Output]

**Already tried**: [List failed approaches to avoid rediscovery]

**Success looks like**: [Testable criterion]

**Respond with**:
- Hypothesis: [Your proposed mechanism in one sentence]
- Reasoning: [Why your domain expertise suggests this]
- Confidence: [low/medium/high]
- Testable prediction: [If my hypothesis is correct, then X should be true]

得:要自足——臣得此文即有思之所需。

敗則:不能述五例或驗法者,疾未備為多臣之問。先縮其範。

第二步:謀其波

列諸可得之臣,分為約十之波。前二波之序無關;後波之間,波際知識之注益其果。

# List all agents from registry
grep '  - id: ' agents/_registry.yml | sed 's/.*- id: //' | shuf

授臣於波。初謀四波——或不需盡用(見第四步之早止)。

要之變
1-220 臣標準之要
310 臣 + advocatus-diaboli要 + 新興共識 + 對辯之挑
4+各 10 臣要 + 「X 已驗。專注邊例與敗。」

得:波授之表,諸臣皆有所屬。納 advocatus-diaboli 於第三波(非後),俾對辯之過影後波。

敗則:可得之臣少於 20 者,減為二三波。十臣亦可,唯合之信稍弱。

第三步:發其波

每波並發為臣。用 sonnet 模以省(其值在見之多元,非各深)。

法甲:TeamCreate(推為全縱)

用 Claude Code 之 TeamCreate 立有任之追之合作團。TeamCreate 為延遲之具——先以 ToolSearch("select:TeamCreate") 取之。

  1. 立團:
    TeamCreate({ team_name: "unleash-wave-1", description: "Wave 1: open-ended hypothesis generation" })
    
  2. 每臣以 TaskCreate 立一任,含要與域之框
  3. Agent 具發每臣為團友,team_name: "unleash-wave-1"subagent_type 設為臣之類(如 kabalistgeometrist
  4. TaskUpdateowner 授任於團友
  5. TaskList 監進——團友自畢自記之
  6. 波間,以 SendMessage({ type: "shutdown_request" }) 閉當前團,立次團而更要(第四步)

此給內合作:共任列追何臣已應,團友可訊以續,領以任授管理波之轉。

法乙:原 Agent 之發(簡,為小行)

每波之臣以要與域之框發之:

Use the [agent-name] agent to analyze this problem through your domain expertise.
[Paste the brief]
Think about this from your specific perspective as a [agent-description].
[For non-technical agents: add a domain-specific framing, e.g., "What patterns
does your tradition recognize in systems that exhibit this kind of threshold behavior?"]
Respond exactly in the requested format.

一波之諸臣以 Agent 具並發,run_in_background: true。俟波畢方發次波(俾第四步之波際知識注)。

二法之擇

TeamCreate原 Agent
宜於第三層全縱(40+ 臣)第二層板(5-10 臣)
合作任列、訊、屬發後不顧、手聚
波際交接任狀承須手追
開銷高(每波設團)低(每臣一具呼)

得:每波二至五分內返約十結構之應。臣不應或出格者記之而不阻管線。

敗則:一波過半敗者,察要之清。常因:出之模歧,或例不足以使外域之臣思之。

第四步:注波際知識(並評早止)

一二波後,發次波之前,取其新興之信。

  1. 掃已畢諸波之應,求重現之題
  2. 識最常之假設族(合之信)
  3. 察早止之閾:頂族二十臣後已逾空模期之三倍者,信強矣。謀第三波為對辯+精煉之波,並考其後止之
  4. 為次波更其要:
**Update from prior waves**: [N] agents independently proposed [hypothesis family].
Build on this — what explains the remaining cases where this hypothesis fails?
Do NOT simply restate this finding. Extend, challenge, or refine it.

早止之囑:非每縱皆需諸臣。已定之疾域(如碼庫之析)合常於 30-40 臣穩。抽象或開放之疾(如未知數學之變),全名冊益,蓋正域實不可預也。每波後察合——頂族之數與空模比已平者,續波之益遞減。

此免重發(後波獨立再得早波所得),導後臣於疾之邊。

得:後波生更精細、有針之假設,補新興共識之缺。

敗則:二波後無合者,疾或過寬。考縮其範或多備例。

第五步:聚而去重

諸波畢後,聚諸應於一文。以族聚假設而去重:

  1. 取諸假設之述
  2. 以機制聚(非以辭——「mod 94 之模算」與「Z_94 之循環群」乃同族)
  3. 計每族之獨立發現之數
  4. 依合排之:更多臣獨立發現之族居高

得:以合計排之假設族列,含貢臣與代之可試之測。

敗則:每假設皆獨(無合)者,信噪比過低。或疾需多例,或臣需更緊之出之式。

第六步:對空模而驗

試頂假設於空模以確合有意,非共訓之偽。

  • 程驗:假設出可試之式或法者,於留出之例行之
  • 空模:估 N 臣偶合於同假設族之概(如有 K 合理族,隨機合概約 N/K)
  • :合逾空模期之三倍者,信為有意

得:頂假設族顯逾偶然之合,並/或過程驗。

敗則:頂假設驗敗者,察次族。皆敗者,疾或需異法(更深單家析、多數、或重設例)。

第七步:對辯之精煉

宜於第三波,非合成之後。 第三波納 advocatus-diaboli(與波際知識注並)較獨立後對辯之過更效。早挑使第四+波對挑而精,非堆於未挑之共識上。

第三波已含對辯者,此步為終察。否(如諸波無之)發 advocatus-diaboli(或 senior-researcher)今。為結構之過,用 TeamCreate 立評團,二臣並對共識:

Here is the consensus hypothesis from [N] independent agents:
[Hypothesis]
[Supporting evidence and convergence stats]

Your job: find the strongest counterarguments. Where does this fail?
What alternative explanations are equally consistent with the evidence?
What experiment would definitively falsify this hypothesis?

得:諸反論、邊例、與證偽之實。假設過對辯之察者,可入合矣。良對辯之過或部分護共識——識其設較他更佳雖不完。

敗則:對辯臣得致命之缺者,反饋以入針之續波(第三層+迭模——擇五至十臣最宜對特挑者)。

第八步:交於團

縱發疾;團解之。化已驗之假設族為可行之事,立焦團解之。

  1. 每已驗假設族立 GitHub 事(用 create-github-issues 之術)
  2. 依合強與影排之
  3. 每事以 TeamCreate 立小團:
    • teams/ 中已定之團合疾域者用之
    • 無合者,默用 opaque-team(N 個 shapeshifter,適性授角)——其能應未知形而不需訂製
    • 至少一非技術之臣(如 advocatus-diabolicontemplative)——其捕技術臣所漏之施險
    • 階段間用 REST 點以免急促
  4. 管線為:縱 → 分類 → 每事一團 → 解

得:每假設族對一追之事,有團授之。縱生診斷;團生修。

敗則:團組與疾不合者,重授。Shapeshifter 之臣可研可設而無寫之具——團領須施其碼之囑。

  • 諸可得之臣已問(或有理之子集已擇)
  • 應以結構可解之式聚
  • 假設已去重,依獨立合排
  • 頂假設已對空模或程試驗
  • 對辯之過挑共識
  • 終假設含可試之測與已知之限

  • 要中例少:臣需五以上之例方尋形。三例則諸臣多以表面對形或模回(以異辭重述要)
  • 無驗之路:無試假設之法者,不能辨信於噪。合僅必要而非足
  • 隱喻之應:域家之臣(mystic、shaman、kabalist)或以富隱喻之理應之,難以程解。出之模納「以可試之式或法述假設」
  • 波間之重發:無波際知識注者,三至七波獨立重發一二波所得。常於波間更要
  • 過解合:43% 合於機制族似深,然察基率。若僅三合理機制族,隨機合約 33%
  • 期單族獨大:抽象之疾(形識、密術)多生一獨大假設族。多維之疾(碼庫析、系設)生諸有效族之廣合——此為期且健,非模之敗
  • 非技術臣之泛框:非技術臣之貢之質依要如何以其域之語框疾。「於此閾你之傳云何系?」生結構之識;泛要無生。投於外域臣之域之框
  • 用此為執行:此模生假設,非施。已驗之假設既得,化為事而交團(第八步)。管線為縱 → 分類 → 每事一團

  • forage-solutions — 蟻群最佳化以探解空(互補:較窄範、較深探)
  • build-coherence — 蜂民主以擇諸競法(用此術後以擇頂假設)
  • coordinate-reasoning — 痕跡合作以管臣間信流
  • coordinate-swarm — 廣群合作之模於分散系
  • expand-awareness — 縮前先放(互補:用為個臣之備)
  • meditate — 發前清脈絡之噪(推於第一步前)

GitHub Repository

pjt222/agent-almanac
Path: i18n/wenyan/skills/unleash-the-agents
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Related Skills

content-collections

Meta

This 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.

View skill

polymarket

Meta

This 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.

View skill

creating-opencode-plugins

Meta

This 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.

View skill

sglang

Meta

SGLang 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.

View skill