Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add AgenticAIPlan/AgenticAISkills --skill llm-data-researchgit clone --depth 1 https://github.com/AgenticAIPlan/AgenticAISkillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/agenticaiplan/agenticaiskills/llm-data-research)<a href="https://agentmods.dev/skills/agenticaiplan/agenticaiskills/llm-data-research"><img src="https://agentmods.dev/badge/skills/agenticaiplan/agenticaiskills/llm-data-research/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/agenticaiplan/agenticaiskills/llm-data-research"><img src="https://agentmods.dev/badge/skills/agenticaiplan/agenticaiskills/llm-data-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00471 | $0.05882 |
| Opus 5 | $0.00235 | $0.02941 |
| Sonnet 5 | $0.00094 | $0.01176 |
| Haiku 4.5 | $0.00047 | $0.00588 |
Grade D, and why
llm-data-research scanned grade D with 3 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
curl -s -X POST "http://localhost:3456/click?target=$XHS_ID" -d '.close-icon' > /dev/null Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -s http://localhost:3456/targets | python3.11 -c "import json,sys; t=json.load(sys.stdin); print(f'CDP就绪,{len(t)}个Tab')" Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s http://localhost:3456/targets | python3.11 -c "import json,sys; t=json.load(sys.stdin); print(f'CDP就绪,{len(t)}个Tab')" How it starts
The opening of the file, as written. The whole thing — 478 lines — stays where its author put it; the contents beside it link to each section on GitHub.
大模型数据动态定期洞察 Skill
本 Skill 自动执行完整的大模型数据市场调研流程,涵盖数据采集、Excel 整理、合作线索关联、市场洞察报告输出、定时任务设置、Word 报告生成共六个步骤。
新增核心能力:监控头部厂商和外部数商的招募动态(数据标注任务招募、专家招募画像),通过招募信号反推:
- 头部大厂下一代模型的发力方向(招募什么样的专家 → 在训练哪个能力方向)
- 外部数商(Talents AI、一面千识等)的生产能力和专家储备质量
- 海外大厂(OpenAI、Google、Meta 等)的数据战略动向
环境前置检查
在开始前确认以下环境就绪:
# 1. 确认 CDP Proxy 运行(web-access skill 依赖)
curl -s http://localhost:3456/targets | python3.11 -c "import json,sys; t=json.load(sys.stdin); print(f'CDP就绪,{len(t)}个Tab')"
# 2. 确认 Python 版本(openpyxl 安装在 python3.11)
python3.11 -c "import openpyxl; print('openpyxl', openpyxl.__version__)"
若 CDP Proxy 未启动,运行:node ~/.claude/skills/web-access/scripts/check-deps.mjs
关键约束:
- 所有浏览器操作必须在主线程用 Bash 直接执行,不得启动子 Agent 执行网络操作(子 Agent 无法继承工具权限)
- Excel 生成使用
python3.11(不是 python3 或 python) - 小红书详情页需登录才能看正文,搜索结果列表页关闭弹窗后可读取卡片信息
工具受限时的降级策略
如果 Bash/Write 工具被拒绝,不要停止,切换到「知识合成 + 内联输出」模式继续执行所有6步:
- 继续完成所有6步,文件输出改为内联 Markdown 输出,并在每步开头注明:
> ⚠️ 工具受限:以下为内联输出,建议手动保存 - Step 1:列出将执行的具体搜索关键词,从训练知识中综合近期信息(明确标注「来源:训练知识,截止至 XXXX 年」)
- Step 2 Excel:输出各 Sheet 的列名 + 至少5条真实数据行(Markdown 表格格式)
- Step 4 报告:完整输出所有8节内容,各节有实质分析,不留空节
- Step 6 Word:输出完整
/tmp/llm_report_data.jsonJSON 内容(代码块包裹),提示用户可手动运行node scripts/generate_report.js --data /tmp/llm_report_data.json
用户真正需要的是分析质量,不只是文件存在。在工具受限时,高质量的内联分析报告比「无法执行」更有价值。
Step 1:多平台信息采集
1.1 小红书采集
小红书有反爬机制,必须通过 CDP 操作:
# 打开搜索页
XHS_ID=$(curl -s "http://localhost:3456/new?url=https://www.xiaohongshu.com/search_result?keyword=大模型专家标注招募" | python3.11 -c "import json,sys; print(json.load(sys.stdin)['targetId'])")
sleep 4
# 关闭登录弹窗(弹窗选择器)
curl -s -X POST "http://localhost:3456/click?target=$XHS_ID" -d '.close-icon' > /dev/null
sleep 1
# 提取搜索结果卡片
curl -s -X POST "http://localhost:3456/eval?target=$XHS_ID" -d '
JSON.stringify(Array.from(document.querySelectorAll("section.note-item, [class*=\"note-item\"]")).map(el => {
const title = el.querySelector("a[class*=\"title\"] span, [class*=\"title\"] span")?.innerText?.trim() || "";
const author = el.querySelector("[class*=\"author\"] span")?.innerText?.trim() || "";
const time_text = el.innerText?.match(/\d{4}-\d{2}-\d{2}|\d+天前|\d+小时前|[0-9]{2}-[0-9]{2}/)?.[0] || "";
const href = el.querySelector("a[href*=\"explore\"]")?.href || "";
return {title, author, time_text, url: href};
}).filter(x=>x.title).slice(0,20))
'
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 478 lines · 471 tokens per session scan D 5bb04ba6f63a
llm-data-research is a skill published in the GitHub repository AgenticAIPlan/AgenticAISkills (11 stars, last pushed 3mo ago), licensed MIT. It adds 471 tokens to every session and 5,882 once invoked, about $0.0024 per session on Opus 5. A static security scan graded it D with 3 findings (sends data to an external url, downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
dcf-model
Build discounted cash flow valuation workbooks in Excel.
audit-xls
Audit a spreadsheet for formula accuracy, errors, and common mistakes. Scopes to a selected range, a single sheet, or the entire model (including financial-model integrity checks like BS balance, cash tie-out, and logic sanity). Triggers on "audit this sheet", "check my formulas", "find formula errors", "QA this…
google-drive-sheets
Find, read, export, edit, and manage the user's Google Drive, Docs, Sheets, and Slides through per-user OAuth.
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
excel-basic-statistics-and-routing
An Excel workflow for filtering grouped data, calculating averages, extracting row ranges, removing duplicates, and adding totals.
dynamic-percentage-and-large-file-analysis
An Excel analysis workflow that changes its file-processing approach based on file size and calculates selected values and percentages.