kol-pricing

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

KOL Pricing

KOL定价

Overview

概述

Analyze X/Twitter KOLs as an agent workflow instead of a local click-through app. The agent fetches public profile and tweet data through UnifAPI MCP tools when available, applies the KOL Pricing framework deterministically, then produces campaign-ready pricing, ROI, warnings, and outreach briefs.

Core principle: keep the original framework's tiering and pricing logic as the source of truth. Use the calling agent's model for synthesis and DM copy; do not require a separate LLM provider key for the core analysis.

Attribution: this skill is adapted from Antoniaiaiaiaia/kol-pricing, originally by Antonia (@antoniayly). See references/original-license.md.

以Agent工作流而非本地点击式应用的方式分析X/Twitter KOL。当可用时，Agent会通过UnifAPI MCP工具获取公开资料和推文数据，确定性地应用KOL定价框架，然后生成可用于营销活动的定价、ROI、风险提示和触达简报。

核心原则：将原框架的分级和定价逻辑作为事实依据。使用调用该技能的Agent模型进行内容合成和DM文案撰写；核心分析不需要单独的LLM提供商密钥。

来源说明：本技能改编自Antoniaiaiaiaia/kol-pricing，最初由Antonia (@antoniayly)开发。详见references/original-license.md。

References

参考资料

references/pricing-logic.md - tier matrix, boosts, penalties, ROI formula, top-pick rules, warnings, and DM boundary.
references/original-license.md - original MIT license notice and attribution.
../unifapi/references/twitter-x.md - current UnifAPI X/Twitter route map.

references/pricing-logic.md - 分级矩阵、加分项、减分项、ROI公式、首选规则、风险提示和DM沟通边界。
references/original-license.md - 原始MIT许可声明及来源说明。
../unifapi/references/twitter-x.md - 当前UnifAPI X/Twitter路由映射。

Workflow

工作流

Gather campaign context.
- Product name, URL, pitch, desired action, estimated LTV.
- Target KOL tiers, excluded tiers, follower floor, engagement floor, extra keywords.
- Handles to analyze, or a search query if discovery is needed.
- If the user does not provide context, ask for minimal campaign inputs. Do not assume the original app's local config files exist in this skills repository.
Fetch public X/Twitter data.
- Prefer the available UnifAPI MCP tools. Look for operations corresponding to:
  - ```
  GET /x/users/by/username/{username}
```
  for profile lookup by handle.
- ```
GET /x/users/{id}/tweets
```
    for recent authored posts after resolving the handle to
```
data.id
```
    .
  - ```
  GET /x/tweets/search/recent
```
  and
```
  GET /x/autocomplete
```
  for candidate discovery.
- For each handle, fetch one profile and recent authored tweets. Ten tweets is enough for the default pricing workflow unless the user asks for deeper evidence.
- Read follower and engagement metrics from
```
public_metrics
```
  , not old flat fields.
- Keep the returned
```
billing
```
  metadata when available so final reports can mention actual record cost.
- Do not call
```
api.x.com
```
  directly unless the user explicitly asks for an official X implementation.
Create a snapshot JSON for deterministic analysis.
- Use this shape:

json

{
  "product": {
    "name": "YourProduct",
    "tagline": "What you do, in one line.",
    "pitch": "Short product pitch.",
    "desired_action": "sign up",
    "ltv_usd": 120,
    "twitter_handle": "@yourhandle",
    "url": "https://example.com"
  },
  "ideal_kols": {
    "preferred_tiers": ["T", "B"],
    "excluded_tiers": [],
    "extra_keywords": ["sdk", "agent"],
    "min_followers": 1000,
    "engagement_floor_pct": 0.5
  },
  "handles": [
    {
      "handle": "example",
      "profile": { "...": "UnifAPI X user object from response.data" },
      "tweets": [{ "...": "UnifAPI X tweet object from response.data[]" }]
    }
  ]
}

The analyzer also accepts whole UnifAPI response envelopes as
```
profile_response
```
and
```
tweets_response
```
, which is useful when preserving
```
request_id
```
,
```
pagination
```
, and
```
billing
```
beside the normalized report.

Run the offline pricing script when a reproducible artifact is useful.

bash

node skills/kol-pricing/scripts/analyze-snapshot.mjs \
  --input /tmp/kol-pricing-input.json \
  --out /tmp/kol-pricing-report.md \
  --json /tmp/kol-pricing-report.json

Draft outreach with the calling agent, not an external LLM key.
- Use
```
dm_brief
```
  from the JSON report.
- Reference exactly one recent tweet when possible.
- Keep the tone practitioner, direct, and low-hype.
- If the recommendation is
```
skip
```
  , draft a zero-cash affiliate/gift-access option only if the user still wants outreach.

收集营销活动背景信息。
- 产品名称、URL、宣传语、期望用户行动、预估LTV。
- 目标KOL等级、排除的等级、粉丝下限、互动率下限、额外关键词。
- 待分析的账号ID，或用于发现候选KOL的搜索查询。
- 如果用户未提供背景信息，请求获取最少的营销活动输入信息。不要假设本技能仓库中存在原应用的本地配置文件。
获取X/Twitter公开数据。
- 优先使用可用的UnifAPI MCP工具。寻找对应以下操作的接口：
  - ```
  GET /x/users/by/username/{username}
```
  ：通过账号ID查找用户资料。
- ```
GET /x/users/{id}/tweets
```
    ：将账号ID解析为
```
data.id
```
    后，获取该用户近期发布的推文。
  - ```
  GET /x/tweets/search/recent
```
  和
```
  GET /x/autocomplete
```
  ：用于发现候选KOL。
- 对于每个账号ID，获取一份用户资料和近期发布的推文。默认定价工作流中，10条推文足够，除非用户要求更深入的证据。
- 从
```
public_metrics
```
  中读取粉丝数和互动指标，而非旧的扁平字段。
- 若有返回的
```
billing
```
  元数据，请保留，以便最终报告中提及实际记录成本。
- 除非用户明确要求官方X平台实现，否则不要直接调用
```
api.x.com
```
  。
创建用于确定性分析的快照JSON。
- 使用如下结构：

json

{
  "product": {
    "name": "YourProduct",
    "tagline": "What you do, in one line.",
    "pitch": "Short product pitch.",
    "desired_action": "sign up",
    "ltv_usd": 120,
    "twitter_handle": "@yourhandle",
    "url": "https://example.com"
  },
  "ideal_kols": {
    "preferred_tiers": ["T", "B"],
    "excluded_tiers": [],
    "extra_keywords": ["sdk", "agent"],
    "min_followers": 1000,
    "engagement_floor_pct": 0.5
  },
  "handles": [
    {
      "handle": "example",
      "profile": { "...": "UnifAPI X user object from response.data" },
      "tweets": [{ "...": "UnifAPI X tweet object from response.data[]" }]
    }
  ]
}

分析器也接受完整的UnifAPI响应包作为
```
profile_response
```
和
```
tweets_response
```
，这在保留
```
request_id
```
、
```
pagination
```
和
```
billing
```
并与标准化报告一起展示时非常有用。

当需要可复现的成果时，运行离线定价脚本。

bash

node skills/kol-pricing/scripts/analyze-snapshot.mjs \
  --input /tmp/kol-pricing-input.json \
  --out /tmp/kol-pricing-report.md \
  --json /tmp/kol-pricing-report.json

使用调用该技能的Agent撰写触达文案，无需外部LLM密钥。
- 使用JSON报告中的
```
dm_brief
```
  。
- 尽可能精确引用一条近期推文。
- 保持语气专业、直接、不过度宣传。
- 如果建议为
```
skip
```
  ，仅当用户仍希望进行触达时，才起草零现金的联盟合作/赠品访问选项。

Output

输出内容

For single-handle analysis, return:

Verdict: tier, top pick, cash range, ROI, risk level.
Evidence: matched keywords, engagement, profile fit, recent tweet signals.
Recommendation: contract terms and outreach brief.
Cost: UnifAPI records consumed or the best estimate if billing metadata is unavailable.

For batch analysis, return:

Ranked table.
Per-KOL mini reports.
Top 3 actions: engage, negotiate, skip.
Optional DM drafts for only the selected KOLs unless the user asks for all.

对于单账号ID分析，返回：

结论：等级、首选标识、现金价格范围、ROI、风险等级。
依据：匹配的关键词、互动率、资料契合度、近期推文信号。
建议：合同条款和触达简报。
成本：消耗的UnifAPI记录数，或在无账单元数据时的最佳估算值。

对于批量分析，返回：

排名表格。
每个KOL的迷你报告。
三大行动建议：接洽、协商、跳过。
可选：仅为选定的KOL起草DM文案，除非用户要求为所有KOL撰写。

Pricing Logic

定价逻辑

Read

references/pricing-logic.md

before changing constants, explaining the model, or reviewing pricing behavior. It records the tier matrix, boosts, penalties, ROI formula, top-pick rules, warnings, and DM boundary from the original code.

在修改常量、解释模型或审核定价行为前，请阅读

references/pricing-logic.md

。该文档记录了原始代码中的分级矩阵、加分项、减分项、ROI公式、首选规则、风险提示和DM沟通边界。

Guardrails

约束规则

Be clear that pricing is a decision aid, not a guaranteed market rate.
Do not hide low-confidence inputs. If tweets are unavailable, protected, too old, or too few, report that.
Do not require
```
ANTHROPIC_API_KEY
```
; the agent using this skill can draft copy itself.
Preserve author attribution when presenting this as an extension of the KOL Pricing framework.

明确说明定价仅为决策辅助工具，并非保证的市场价格。
不要隐藏低置信度输入。如果推文不可用、受保护、过旧或数量过少，请如实报告。
不需要
```
ANTHROPIC_API_KEY
```
；使用本技能的Agent可自行撰写文案。
当将此作为KOL定价框架的扩展展示时，请保留作者署名。