/ 目錄 / 演練場 / infinite-gratitude
● 社群 sstklen ⚡ 即開即用

infinite-gratitude

作者 sstklen · sstklen/infinite-gratitude

10 specialized agents collaborate on one research question — a multi-agent research dojo where each agent plays a distinct role.

Part of the 'AI Dojo' series. Spawns 10 agents with different roles (explorer, critic, synthesizer, etc.) that collaborate on a research task. Each agent commits its piece and the skill coordinates the handoffs — designed for tough research questions where one-shot prompting falls flat.

為什麼要用

核心特性

即時演示

實際使用效果

infinite-gratitude-skill.replay ▶ 就緒
0/0

安裝

選擇你的客戶端

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "infinite-gratitude-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/sstklen/infinite-gratitude",
        "~/.claude/skills/infinite-gratitude"
      ],
      "_inferred": true
    }
  }
}

開啟 Claude Desktop → Settings → Developer → Edit Config。儲存後重啟應用。

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "infinite-gratitude-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/sstklen/infinite-gratitude",
        "~/.claude/skills/infinite-gratitude"
      ],
      "_inferred": true
    }
  }
}

Cursor 使用與 Claude Desktop 相同的 mcpServers 格式。專案級設定優先於全域。

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "infinite-gratitude-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/sstklen/infinite-gratitude",
        "~/.claude/skills/infinite-gratitude"
      ],
      "_inferred": true
    }
  }
}

點擊 Cline 側欄中的 MCP Servers 圖示,然後選 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "infinite-gratitude-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/sstklen/infinite-gratitude",
        "~/.claude/skills/infinite-gratitude"
      ],
      "_inferred": true
    }
  }
}

格式與 Claude Desktop 相同。重啟 Windsurf 生效。

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "infinite-gratitude-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/sstklen/infinite-gratitude",
        "~/.claude/skills/infinite-gratitude"
      ]
    }
  ]
}

Continue 使用伺服器物件陣列,而非映射。

~/.config/zed/settings.json
{
  "context_servers": {
    "infinite-gratitude-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/sstklen/infinite-gratitude",
          "~/.claude/skills/infinite-gratitude"
        ]
      }
    }
  }
}

加入 context_servers。Zed 儲存後熱重載。

claude mcp add infinite-gratitude-skill -- git clone https://github.com/sstklen/infinite-gratitude ~/.claude/skills/infinite-gratitude

一行命令搞定。用 claude mcp list 驗證,claude mcp remove 移除。

使用場景

實戰用法: infinite-gratitude

Tackle a research question single-agent prompts fail on

👤 Researchers, analysts, strategists ⏱ ~60 min advanced

何時使用: Questions that need breadth (many sources) and depth (skepticism, synthesis) — where one pass hallucinates.

前置條件
  • Skill installed — git clone https://github.com/sstklen/infinite-gratitude ~/.claude/skills/infinite-gratitude
步驟
  1. Pose the question
    Use infinite-gratitude. Research question: what are the real bottlenecks in scaling post-transformer architectures in 2026? Run the 10-agent dojo.✓ 已複製
    → Roles spin up: explorer, critic, synthesizer, etc., each producing a distinct contribution
  2. Watch the handoffs
    Show me the critic's pushback on the explorer's output.✓ 已複製
    → Substantive critique, not agreement
  3. Final synthesis
    Give me the synthesizer's final artifact with dissent noted.✓ 已複製
    → Document with main thesis + minority views preserved

結果: A research brief richer than a single-agent answer.

注意事項
  • Agents converge and critic rubber-stamps — Prompt the critic explicitly to find flaws; force disagreement
搭配使用: filesystem

Draft a policy memo with explicit dissenting views

👤 Policy wonks, internal strategy teams ⏱ ~90 min advanced

何時使用: The decision needs to survive scrutiny from people who'll disagree.

步驟
  1. Frame the memo
    Use infinite-gratitude. Memo topic: should our engineering org adopt four-day work weeks? Run the full dojo.✓ 已複製
    → Multi-perspective outputs on pros, cons, empirical evidence, implementation risks
  2. Extract the dissent
    Keep the critic's dissent intact as an appendix; don't blend it away.✓ 已複製
    → Main memo + appendix of unresolved concerns

結果: A memo your skeptics can see was genuinely considered.

組合

與其他 MCP 搭配,撬動十倍槓桿

infinite-gratitude-skill + paper-fetch-skill

Explorer agent downloads papers; synthesizer cites them

Run the dojo on [topic] and use paper-fetch-skill for the explorer's sources.✓ 已複製
infinite-gratitude-skill + filesystem

Save each agent's output as its own file for traceability

Persist each agent's contribution to research/session-<date>/<role>.md✓ 已複製

工具

此 MCP 暴露的能力

工具輸入參數何時呼叫成本
dojo_run question, optional agent roster Big, ambiguous questions 0 (but high token use)
single_agent role name, task Just one role needed 0

成本與限制

運行它的成本

API 配額
None
每次呼叫 Token 數
High — 10 agents each doing real work
費用
Free (Claude API usage applies)
提示
Use for high-stakes questions only. Single-agent suffices for most research.

安全

權限、密鑰、影響範圍

憑證儲存: None
資料出站: None from skill itself

故障排查

常見錯誤與修復

Dojo feels repetitive

Agents may be blending; enforce distinct perspectives in prompts

Runs out of context mid-session

Save intermediate outputs to disk and resume fresh; don't carry everything in one session

替代方案

infinite-gratitude 對比其他方案

替代方案何時用它替代權衡
Single-agent with explicit stepsStraightforward questionsLess depth, cheaper
planning-with-files-skillYou want durable state across long sessions but not multi-roleNo role specialization

更多

資源

📖 閱讀 GitHub 上的官方 README

🐙 查看未解決的 issue

🔍 瀏覽全部 400+ MCP 伺服器和 Skills