Description

数据自动分析 + 广告投放优化一体化 Skill。当用户上传 Excel/CSV 文件，或提到以下任一场景时必须触发：①通用数据分析（看报表、数据趋势、可视化）；②账户诊断（哪些计划效果差、哪些要暂停、投放诊断、账户体检）；③A/B 测试分析（两组数据对比、哪个版本好、是否显著、置信度）；④日报生成（投放日报、每...

README (SKILL.md)

数据自动分析 Skill

Name: 数据自动分析
Author: ming0429

一体化数据分析与广告优化工具集，包含 4 个模式，根据用户意图选择对应模式执行。

环境准备（所有模式通用）

python3 -m venv /home/claude/.venv && source /home/claude/.venv/bin/activate
pip install pandas openpyxl xlrd jinja2 scipy -q

模式选择决策树

用户上传了 Excel/CSV？
├── 是
│   ├── 提到"诊断/体检/哪些要暂停/计划效果差" → 模式 B：账户诊断
│   ├── 提到"日报/汇报/对比昨日" → 模式 D：日报生成
│   ├── 提到"A/B 测试/显著性/哪个版本好" → 模式 C：A/B 测试
│   └── 其他（看报表/分析数据/趋势） → 模式 A：通用分析
└── 否
    └── 用户直接描述两组数据（手工输入） → 模式 C：A/B 测试

不确定时优先询问用户，不要猜。

模式 A：通用数据分析

用途：任意 Excel/CSV 都能用，自动识别列类型，生成交互式 HTML 报告。

python3 scripts/analyze.py --file \x3C输入文件> --out /mnt/user-data/outputs/data_report.html

输出：包含数据概览、指标汇总、异常检测、可分页表格、5 个 ECharts 图表的 HTML。

模式 B：广告账户诊断器

用途：分析广告投放报表，给每条计划打红/黄/绿预警，输出处置建议（暂停/降价/提价/观察）。

python3 scripts/diagnose.py --file \x3C投放报表.xlsx> --out /mnt/user-data/outputs/diagnose_report.html

诊断规则摘要（完整说明见 references/diagnose_rules.md）：

🔴 红色（立即处理）：消耗 > 均值 2 倍但转化为 0、CPA > 均值 3 倍、CTR \x3C 均值 0.3 倍
🟡 黄色（需优化）：CPA > 均值 1.5 倍、转化率 \x3C 均值 0.5 倍
🟢 绿色（健康）：指标正常或优于均值

脚本自动识别常见列名（消耗/花费/cost、转化/conversion、点击/click、展示/impression 等），识别失败时提示用户手动指定。

模式 C：A/B 测试结果分析

用途：判断两组数据差异是否显著，给出置信度和结论。

两种场景：

比例型（CTR、转化率）→ Z 检验
均值型（CPC、CPA、ROI）→ T 检验

执行方式 1：手工输入数据

python3 scripts/ab_test.py --inline \
  --a-success 120 --a-total 2400 \
  --b-success 150 --b-total 2500 \
  --out /mnt/user-data/outputs/ab_result.html

执行方式 2：从文件

python3 scripts/ab_test.py --file \x3C数据.xlsx> \
  --group-col \x3C分组列名> --metric-col \x3C指标列名> \
  --metric-type \x3Crate|mean> \
  --out /mnt/user-data/outputs/ab_result.html

详细用法和场景示例见 references/ab_test_guide.md。

模式 D：每日投放日报生成器

用途：上传投放报表，生成结构化日报。输出钉钉/飞书可直接粘贴的纯文本版 + HTML 精美版。

# 单日报表
python3 scripts/daily_report.py --today \x3C今日.xlsx> --out-dir /mnt/user-data/outputs/

# 带昨日对比（推荐）
python3 scripts/daily_report.py --today \x3C今日.xlsx> --yesterday \x3C昨日.xlsx> --out-dir /mnt/user-data/outputs/

输出两个文件：

daily_report.txt — 纯文本，直接复制到钉钉/飞书/微信
daily_report.html — 精美 HTML 版，适合邮件附件或存档

日报结构：核心指标卡片 + 环比涨跌 + TOP 3 最好/最差计划 + 异常提醒 + 明日建议。

详见 references/daily_report_format.md。

Notes

分析过程完全本地执行，不上传任何数据；生成的 HTML 报告在浏览器打开时会从 CDN (cdnjs.cloudflare.com) 加载 ECharts
所有脚本必须保存为 .py 文件执行，不支持 python3 -c 内联
列名完全动态识别，不预设字段名
编码自动识别，兼容 UTF-8 / GBK / GB2312

Usage Guidance

This skill appears internally consistent for local analysis of uploaded Excel/CSV files: it needs Python and a few common packages and contains scripts that operate on user-provided data to produce local HTML/text reports. Before installing/run it: (1) review the included scripts yourself (they are provided) and confirm you are comfortable running third‑party Python code; (2) run pip installs inside an isolated environment (the SKILL.md suggests a venv — you can choose a different path); (3) if you must avoid any network activity, note that pip will fetch packages from PyPI and the generated HTML loads ECharts from a CDN when opened in a browser; (4) ensure output paths (/mnt/user-data/outputs or the venv path) are acceptable for your environment and do not overwrite important files. If you need higher assurance, run the scripts in a sandbox or container and inspect outputs before sharing them.

Capability Analysis

Type: OpenClaw Skill Name: data-auto-analyzer Version: 3.0.1 The skill bundle is a legitimate data analysis and advertising optimization toolset. It includes scripts for A/B testing (ab_test.py), account diagnosis (diagnose.py), and report generation (analyze.py, daily_report.py) using standard libraries like pandas and scipy. The code performs all processing locally on user-provided Excel/CSV files and generates HTML reports with ECharts visualizations. There is no evidence of data exfiltration, malicious execution, or harmful prompt injection; the instructions in SKILL.md are strictly aligned with the stated functional purpose.

Capability Assessment

✓ Purpose & Capability

Name/description request Python + common data-science libs (pandas, openpyxl, xlrd, scipy, etc.) and the included scripts implement data analysis, diagnostics, A/B tests and report generation — these requirements are coherent and proportionate to the advertised functionality.

ℹ Instruction Scope

SKILL.md instructs creation/activation of a virtualenv (/home/claude/.venv), pip installing dependencies, and running the provided scripts on user-supplied Excel/CSV files; scripts only read input files and write local HTML/text outputs. Two small points to note: (1) the venv path is hard-coded to /home/claude/.venv which may be surprising on some hosts, and (2) HTML reports reference ECharts via a public CDN (cdnjs.cloudflare.com) which loads in the user's browser — no code in the skill performs outbound network calls during analysis.

ℹ Install Mechanism

There is no packaged installer; runtime instructions instruct pip installing packages from PyPI into a venv. Installing from PyPI is common for Python tools but is a network operation and should be considered a moderate-risk action compared with an instruction-only skill that requires no installs. The skill does not download/extract arbitrary archives from personal servers or use URL shorteners.

✓ Credentials

The skill requests no environment variables, no credentials, and no config paths. It only requires python3 and typical data-analysis packages — these are proportional to the stated tasks.

ℹ Persistence & Privilege

The skill does not request 'always: true' and does not modify other skills. It does create a venv at /home/claude/.venv and writes output files to user-specified output directories (SKILL.md examples use /mnt/user-data/outputs/). Creating a venv and writing reports is normal, but be aware these create persistent files in the agent's filesystem.

Version History

v3.0.1

- 版本号由 3.0.0 升级到 3.0.1 - 无文件内容变更，仅为元数据中的版本号小幅更新 - 功能、接口和使用文档均未做任何调整

v3.0.0

**重大更新：Skill 3.0.0 现已支持一体化广告智能分析，覆盖日报、账户诊断及 A/B 测试等多业务场景。** - 新增四大模式：通用分析、广告账户诊断、A/B 测试分析、每日投放日报。 - 新增广告账户诊断自动预警与优化建议，支持主流广告平台用词自动适配。 - 支持两组数据自动显著性分析，兼容比例型与均值型 A/B 测试。 - 支持日报自动生成，一键输出钉钉/飞书纯文本与 HTML 版日报。 - 新增规则与格式参考文档（ab_test_guide.md、diagnose_rules.md、daily_report_format.md）。 - 新增脚本：diagnose.py、ab_test.py、daily_report.py，功能结构化、便于扩展。

v1.1.1

- 更新版本号为 1.1.1，对齐 Skill 元数据 - 修改依赖安装建议，推荐在虚拟环境中安装相关依赖包 - 使用 pip 安装时去除 --break-system-packages 参数，提升兼容性 - 优化“Notes”部分表述，明确数据本地处理和 ECharts 使用方式 - 调整部分说明文字，使文档指引更加清晰易读

v1.1.0

Major update: Enhanced automated data analysis and reporting. - Now auto-detects column types (date, dimension, numeric) for any uploaded Excel/CSV file. - Generates a fully interactive HTML report with ECharts charts and paginated, sortable tables. - Adds automated anomaly detection and specific optimization suggestions based on data. - Supports all structured tabular data types (ads, sales, finance, ops metrics, e-commerce, and more). - Requires no setup; works with .xlsx/.xls/.csv, auto-detects encoding, and runs locally for data privacy. - Clearly documents usage, features, and example commands in SKILL.md.

Metadata

Slug data-auto-analyzer

Version 3.0.1

License MIT-0

All-time Installs 0

Active Installs 0

Total Versions 4

Frequently Asked Questions

What is 数据自动分析?

数据自动分析 + 广告投放优化一体化 Skill。当用户上传 Excel/CSV 文件，或提到以下任一场景时必须触发：①通用数据分析（看报表、数据趋势、可视化）；②账户诊断（哪些计划效果差、哪些要暂停、投放诊断、账户体检）；③A/B 测试分析（两组数据对比、哪个版本好、是否显著、置信度）；④日报生成（投放日报、每... It is an AI Agent Skill for Claude Code / OpenClaw, with 132 downloads so far.

How do I install 数据自动分析?

Run "/install data-auto-analyzer" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is 数据自动分析 free?

Yes, 数据自动分析 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does 数据自动分析 support?

数据自动分析 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created 数据自动分析?

It is built and maintained by Ming0429 (@ming0429); the current version is v3.0.1.

More Skills

数据自动分析