Model Pricing Sync

v0.1.1

抓取模型厂商与代码工具价格页面，读取 Markdown 后补 JSON 提取文件，并同步到飞书普通电子表格。需要收集模型价格、补全价格 JSON、追踪价格变化或推送 Lark Sheets 时使用。

⭐ 0· 74·0 current·0 all-time

by@sapereaude2014

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for sapereaude2014/model-pricing-sync.

Previewing Install & Setup.

Prompt PreviewInstall & Setup

Install the skill "Model Pricing Sync" (sapereaude2014/model-pricing-sync) from ClawHub.
Skill page: https://clawhub.ai/sapereaude2014/model-pricing-sync
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install model-pricing-sync

ClawHub CLI

Package manager switcher

npx clawhub@latest install model-pricing-sync

Security Scan

VirusTotal

Suspicious

View report →

OpenClaw

Benign

high confidence

✓

Purpose & Capability

The name/description (collect model/tool pricing and sync to Feishu) matches the code and instructions: Playwright-based page collection, JSON/CSV preparation, and a Feishu/Lark CLI-based push. No unrelated cloud credentials or unexpected system access are requested.

✓

Instruction Scope

SKILL.md limits operations to files inside the skill directory (config/, artifacts/, data/) and running the provided Python scripts. It does web scraping of vendor pages and writes artifacts, which is consistent with the described workflow. There is no instruction to read unrelated system files or secrets.

ℹ

Install Mechanism

There is no packaged install spec (instruction-only install), but requirements.txt lists playwright and instructs installing browsers; the Feishu/Lark integration relies on an external CLI (@larksuiteoapi/cli). This is expected for the functionality, but users must run those install steps themselves. No downloads from untrusted URLs or archive extraction are present in the repo.

ℹ

Credentials

The skill declares no required env vars, which is coherent because Feishu access is performed via the lark-cli subprocess (it expects local CLI configuration/auth). However, this means the skill will use the local lark-cli identity provided (via the --as identity flag) and any credentials configured for that CLI. Users should be aware that the skill will call external sites and the Feishu API using the local CLI's account.

✓

Persistence & Privilege

The skill is not always-enabled and does not request system-wide configuration changes. It writes artifacts and a local sheet_state.json under the skill directory (expected for sync state). Autonomous invocation is allowed by default on the platform but that is normal and not flagged by itself.

Assessment

This skill appears coherent for scraping pricing pages and syncing to Feishu. Before installing: 1) Ensure you install Playwright and run 'playwright install chromium' as the repo suggests (these downloads are large and require network access). 2) Install and configure the Lark/Feishu CLI (npm @larksuiteoapi/cli) and authenticate the account you want the skill to write to—the skill calls lark-cli as a subprocess and will act with that CLI's credentials. 3) Run the skill in a controlled environment (not on a machine with unrelated secrets) because it fetches many public vendor pages and will create/modify spreadsheets in the authenticated Feishu account. 4) Review config/vendors.csv and target CSVs to confirm which vendors/plans will be scraped and tracked. 5) If you plan to allow autonomous agent invocation, be aware the skill can push data to Feishu when invoked; if that is undesired, disable autonomous invocation on the platform. Otherwise there are no clear incoherencies or malicious signs in the provided files.

Like a lobster shell, security has layers — review code before you run it.

feishuvk97493nn6xkbjs5qva7hkxx58985dsp9latestvk97493nn6xkbjs5qva7hkxx58985dsp9pricingvk97493nn6xkbjs5qva7hkxx58985dsp9trackingvk97493nn6xkbjs5qva7hkxx58985dsp9

74downloads

0stars

2versions

Updated 5d ago

v0.1.1

MIT-0

model-pricing-sync

按下面流程执行。

路径约定：本文所有路径都相对于本 skill 目录。执行前先定位 skills/model-pricing-sync/，下文用 <skill_dir> 表示。<run_dir> 表示本次运行生成的目录名，通常为 YYYYMMDD。

新增来源约定：如果用户要求新增某厂商的 API 或订阅跟踪，先搜索并确认官方价格页，再把来源和目标条目补到 <skill_dir>/config/ 下对应 CSV。

1. 采集

运行：

python <skill_dir>/scripts/sync_pricing.py --mode collect

作用：

抓取网页来源并保存为 <skill_dir>/artifacts/<run_dir>/collected/pages/*.md
初始化或合并 <skill_dir>/artifacts/<run_dir>/extracted/api_pricing.json
初始化或合并 <skill_dir>/artifacts/<run_dir>/extracted/subscription_plans.json
更新 <skill_dir>/artifacts/latest_run.json
后续步骤都默认处理 <skill_dir>/artifacts/latest_run.json 指向的这个 run_dir

只想跑部分来源时，可传 --vendors anthropic,google。

采集异常只看：

<skill_dir>/artifacts/<run_dir>/collected/collect_issues.csv

如果有记录，先根据 source_url 和 message 判断原因：

临时网络、超时、正文异常、反爬验证：先只对对应 vendor 重新运行采集，例如 --vendors openai
来源配置需要调整：更新 <skill_dir>/config/vendors.csv，再只对对应 vendor 重新运行采集

2. 更新目标并生成待填 JSON

只处理当前 run_dir：

<skill_dir>/artifacts/<run_dir>/collected/pages/*.md

先根据本次采集的 Markdown 更新目标表，再生成待填 JSON。目标表是跟踪对象身份的唯一来源；页面 URL 只在 <skill_dir>/config/vendors.csv 维护：

API 目标表：<skill_dir>/config/target_api_models.csv
订阅目标表：<skill_dir>/config/target_subscription_plans.csv

目标表更新规则：

先整体查看两个目标表，确认当前跟踪范围，再逐个阅读 <skill_dir>/artifacts/<run_dir>/collected/pages/*.md，若 Markdown 正文异常、抽取错位或信息明显不全，可按 vendor_key + region + track_type 到 <skill_dir>/config/vendors.csv 找对应 source_url，并直接打开网页作为依据。
如果 Markdown 里出现目标表没有的新旗舰、新主力、新版本数字更新模型，或正式新套餐，必须先把对应记录加入目标表
新增目标默认 required=true、is_active=true，除非页面明确只是临时预告、不可购买或不计划跟踪
新目标只选当前官方价格表主推项
已有套餐新增月付、年付、促销价等计费周期时，在后续同一套餐对象的 prices[] 里补价格
读 Markdown 后确认目标已下线、官方移除或不再跟踪时，把目标表对应记录改为 is_active=false

目标表确认无误后再运行：

python <skill_dir>/scripts/sync_pricing.py --mode prepare

prepare 会按目标表更新当前 run 的两个待填写文件：

<skill_dir>/artifacts/<run_dir>/extracted/api_pricing.json
<skill_dir>/artifacts/<run_dir>/extracted/subscription_plans.json

它只追加缺失的模型 / 套餐对象，不覆盖已经填写的 prices[]。

3. 填写 JSON

打开并填写当前 run 的两个文件：

<skill_dir>/artifacts/<run_dir>/extracted/api_pricing.json
<skill_dir>/artifacts/<run_dir>/extracted/subscription_plans.json

填写依据：

只根据 <skill_dir>/artifacts/<run_dir>/collected/pages/*.md 里明确出现的信息填写
可参考上一轮已填写的 JSON，保持对象映射、original_price_text 写法、notes 风格和套餐/模型摘要的一致性
不得因为参考上一轮 JSON 而跳过本次 Markdown 核对，尤其是价格和模型
正常按 pages/*.md 填；只有当采集结果明显不可靠时，直接访问对应网页补齐
不猜测页面没有明确写出的价格、能力、周期或缓存费用
找不到就保持空数组、空字符串或不填，不要编造价格
保留 source_url、original_price_text 和必要 notes
不要改 prepare 已经预填的身份字段
如果填 JSON 期间又改了目标表，重新运行 --mode prepare

填写 api_pricing.json：

主要填写顶层 notes 和 prices[]
prices[] 字段：price_dimension、source_price_amount、currency、unit_basis、original_price_text、notes
API 价格统一采用基准价口径：官方公开按量付费、文本 token、标准在线推理价
不用免费层、试用额度、促销价、折扣价、预付包、资源包、专属实例、吞吐量包、批处理价或微调价，除非目标对象本身就是这些计费项
页面同时给原价、折扣价、优惠价时，优先取原价 / 标准价；如果页面只给当前成交价，在 notes 说明
页面按上下文长度、输入长度、输出长度分阶梯时，取页面顺序中的第一个基准阶梯，并把阶梯条件写进 original_price_text；notes 统一写 基准阶梯价
页面有文字、图片、音频、视频等多模态价格时，普通模型目标只取文字 token 价格；其他模态只在目标对象明确需要时填写
价格必须来自目标模型所在的同一标题、段落或表格块；相邻价格表不能借用。无法确认归属时保持 prices[] 为空，并在顶层 notes 写明疑点
每个目标模型按 prices[] 拆价格；常用维度是 input、cached_input、output
不适用缓存的厂商只填 input、output

填写 subscription_plans.json：

主要填写顶层 features、plan_summary、notes 和 prices[]
prices[] 字段：billing_cycle、source_price_amount、currency、unit_basis、seat_rule、original_price_text、notes
月付、季付、年付、促销价等放在同一个套餐对象的多个 prices[] 里
套餐权益优先写入 features[]；如果需要人工整理摘要，再写 plan_summary

推荐词表：

API price_dimension：input、output、cached_input、cached_output、embedding、image、audio、other
API unit_basis：per_1m_tokens、per_1k_tokens、per_image、per_minute、per_month、other
订阅 billing_cycle：monthly、quarterly、annual、空
订阅 unit_basis：per_month、per_user_month、per_year、per_user_year、per_quarter、空
region：优先用配置已有值，不自行发明新口径，除非页面明确出现且现有值无法表达

4. 构建与推送

运行：

python <skill_dir>/scripts/sync_pricing.py --mode build

build 会生成扁平 CSV，并把全量问题写到 <skill_dir>/artifacts/<run_dir>/build_issues.csv。

如果 build_issues.csv 里有能补齐的信息，修正后重新运行 --mode build。如果剩余问题是当前材料无法补齐的来源失败、页面未写价格或目标暂时缺价，保留 issue，继续推送。

推送：

python <skill_dir>/scripts/sync_pricing.py --mode push

push 同步最近一次 build 生成的数据。

Comments

Loading comments...