buyer-eval-skill MCP — Сценарии использования, Установка & Живое демо

Зачем использовать

Ключевые функции

Domain-expert question banks for common B2B categories
Structured interview mode — can converse with vendor AI agents
Evidence-based scoring with per-criterion citations
Multi-vendor comparison matrix
Shell-based tooling, open source

Живое демо

Как выглядит на практике

buyer-eval-skill.replay ▶ готово

0/0

Установка

Выберите клиент

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "buyer-eval-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/salespeak-ai/buyer-eval-skill",
        "~/.claude/skills/buyer-eval-skill"
      ],
      "_inferred": true
    }
  }
}

Откройте Claude Desktop → Settings → Developer → Edit Config. Перезапустите после сохранения.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "buyer-eval-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/salespeak-ai/buyer-eval-skill",
        "~/.claude/skills/buyer-eval-skill"
      ],
      "_inferred": true
    }
  }
}

Cursor использует ту же схему mcpServers, что и Claude Desktop. Конфиг проекта приоритетнее глобального.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "buyer-eval-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/salespeak-ai/buyer-eval-skill",
        "~/.claude/skills/buyer-eval-skill"
      ],
      "_inferred": true
    }
  }
}

Щёлкните значок MCP Servers на боковой панели Cline, затем "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "buyer-eval-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/salespeak-ai/buyer-eval-skill",
        "~/.claude/skills/buyer-eval-skill"
      ],
      "_inferred": true
    }
  }
}

Тот же формат, что и Claude Desktop. Перезапустите Windsurf для применения.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "buyer-eval-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/salespeak-ai/buyer-eval-skill",
        "~/.claude/skills/buyer-eval-skill"
      ]
    }
  ]
}

Continue использует массив объектов серверов, а не map.

~/.config/zed/settings.json

{
  "context_servers": {
    "buyer-eval-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/salespeak-ai/buyer-eval-skill",
          "~/.claude/skills/buyer-eval-skill"
        ]
      }
    }
  }
}

Добавьте в context_servers. Zed перезагружается автоматически.

claude mcp add buyer-eval-skill -- git clone https://github.com/salespeak-ai/buyer-eval-skill ~/.claude/skills/buyer-eval-skill

Однострочная команда. Проверить: claude mcp list. Удалить: claude mcp remove.

Сценарии использования

Реальные сценарии: buyer-eval-skill

How to evaluate three SaaS vendors against your requirements

👤 Buyers, RevOps, procurement leads ⏱ ~120 min intermediate

Когда использовать: You've shortlisted 3 vendors and need to pick one defensibly.

Предварительные требования

Skill cloned — git clone https://github.com/salespeak-ai/buyer-eval-skill ~/.claude/skills/buyer-eval-skill

Поток

Define the rubric

We're buying a customer support platform. Propose a weighted rubric (feature depth, integrations, price, security, support).✓ Скопировано

→ Weighted criteria with rationale
Interview each vendor

Use the skill's question bank to interview vendor A. Their docs / AI agent is at <url>. Record evidence per criterion.✓ Скопировано

→ Per-criterion evidence snippets with sources
Score and compare

Score all three and produce a comparison matrix with a recommendation.✓ Скопировано

→ Matrix + recommendation + named risks

Итог: A defensible vendor pick with evidence, not vibes.

Подводные камни

Rubric favors the vendor you already like — Lock the rubric before seeing demos
Evidence is just vendor marketing claims — Weight third-party sources (docs, G2, case studies) over marketing

Сочетать с: filesystem

Compare RFP responses against your requirements

👤 Procurement teams drowning in RFP responses ⏱ ~90 min intermediate

Когда использовать: You sent an RFP, got 5 responses, and need to rank them.

Поток

Load the responses

Here are 5 RFP response PDFs in rfp/. Extract answers per requirement.✓ Скопировано

→ Per-requirement matrix
Flag gaps and fluff

Where did a vendor answer the wrong question or punt? Flag fluff.✓ Скопировано

→ Honest read of each response
Rank

Rank the responses and write a short memo for the evaluation committee.✓ Скопировано

→ Ranked memo

Итог: A fair, fast RFP review.

Подводные камни

Rewarding the best-written response over the best fit — Score on substance; penalize fluff

Комбинации

Сочетайте с другими MCP — эффект x10

buyer-eval-skill + filesystem

Read PDFs/docs from a local folder during evaluation

Read rfp/responses/ and extract answers per requirement.✓ Скопировано

buyer-eval-skill + google-ai-mode-skill

Cross-check vendor claims against public info

For each vendor claim that isn't in their docs, run a quick web check.✓ Скопировано

Инструменты

Что предоставляет этот MCP

Инструмент	Входные данные	Когда вызывать
rubric-design	domain + must-haves	Start of any evaluation
vendor-interview	vendor source (docs/AI agent)	Per vendor
scoring-and-matrix	evidence per vendor + rubric	After interviews
recommendation-memo	matrix	Final step

Стоимость и лимиты

Во что обходится

Квота API: none beyond LLM
Токенов на вызов: 10–40k per full eval
Деньги: free
Совет: Interview one vendor at a time — keeps context focused and evidence clean

Безопасность

Права, секреты, радиус поражения

Хранение учётных данных: none

Исходящий трафик: only if the skill calls out to vendor agents you point it at

Vendor chats may be recorded on their side — treat as a business conversation

Устранение неполадок

Частые ошибки и исправления

Evidence is shallow

Point Claude at primary docs; fallback to published case studies

Scores feel gamed

Re-lock the rubric and re-run scoring blind to vendor identity if possible

Альтернативы

buyer-eval-skill в сравнении

Альтернатива	Когда использовать	Компромисс
creative-director-skill	Evaluating creative concepts, not B2B software	Different judgment domain

Ещё

Ресурсы

📖 Читать официальный README на GitHub

🐙 Открытые задачи

🔍 Все 400+ MCP-серверов и Skills