CSV/Excel Deep Data Analysis Skill
Automatically extracts statistical features, data quality metrics, and anomalies, generating interactive HTML analysis reports.
The skill instructs the LLM to call execute_skill_script_file and html_interpreter tools, but no explicit permission scoping or user confirmation is mentioned. The script only reads user-provided files, with no network or write operations, which limits risk. However, data flow transparency and sensitive data handling are not addressed. Since main risks are visible but permissions and confirmation are incomplete, a score of 12 is given.
csv_analyzer.py code is complete and logical, but error handling for abnormal inputs (e.g., empty files, malformed formats) is weak, and failure messages may be unclear. No test cases or executed verification are available. According to static calibration, reproducibility cannot be verified, so score is capped at 10. Given the script's plausibility, a score of 7 is assigned.
The skill has a clear use case, supports both English and Chinese trigger keywords, and suits data analysis needs. However, it lacks explicit non-fit boundaries (e.g., large files, non-tabular data) and environment fit is unverified (e.g., reachability from mainland China). Score: 11.
SKILL.md is well-structured with workflow, parameter examples, and file structure, but lacks version history, changelog, and maintenance responsibility. The references file contains only TODO placeholders. Overall readable but incomplete, so score is 8.
Static review cannot confirm output usability, but the skill describes a complete analysis workflow that should produce reports. However, no sample outputs or comparative benefit evidence are present. Per static calibration, capped at 7; considering the description is complete but unverified, a score of 6 is given.
No test evidence is provided; only source code and documentation exist. The repository's test files (e.g., test_dispatch_tool.py) do not directly cover this skill and cannot be used as evidence. Per static calibration, capped at 5; here a score of 3 is assigned.
- The skill does not explicitly require user confirmation or least-privilege principles; verify that tool calls are scoped to necessary file access before use.
- Error feedback for abnormal files (e.g., empty files, encoding issues) may be unclear; consider adding more explicit error logs.
- The skill does not document file size limits, performance constraints, or non-tabular data boundaries, which may affect real-world usage.
- No test cases or executed verification are provided; static review cannot confirm the script's reliability in real scenarios.
- The reference document (reference.md) contains only TODO placeholders, lacking detailed API documentation and workflow examples.
What does this skill do, and when should you use it?
This skill analyzes CSV, Excel, or TSV files and generates interactive HTML reports containing statistical summaries, data quality checks, numerical distributions, correlations, categorical structures, anomaly detection, attribution analysis, and recommendations. It runs a Python script to extract features, then injects them into an ECharts and Tailwind CSS-based template. It supports English and Chinese, emphasizing LLM-driven business insights. As part of the DB-GPT repository, it requires the DB-GPT environment to run.
The skill uses execute_skill_script_file to run scripts/csv_analyzer.py with the input file path. The script returns a statistical summary and marker-wrapped data blocks. Then the html_interpreter tool loads templates/report_template.html and fills in 9 placeholders (language, title, summary, various insights), while chart data is automatically injected from markers by the backend. The output is an interactive HTML report with ECharts charts, divided into foundational analysis and anomaly attribution sections.
- A data analyst needs to quickly understand distributions, correlations, and outliers in a large CSV dataset and produce a visual report for stakeholders.
- A business user uploads Excel sales data and expects a deep analysis report with charts and business recommendations for decision-making.
- A team on the DB-GPT platform wants to generate shareable HTML reports from files, instead of static images.
- Scenarios requiring bilingual reports, e.g., sharing analysis results across international teams.
What are this skill's strengths and limitations?
- Generates interactive HTML reports with multiple charts without manual coding
- Supports CSV, Excel, and TSV formats
- Separates statistical extraction from LLM insights for richer reports
- Supports English and Chinese, auto-detecting user language
- Lightweight: transmits only core JSON data for fast rendering
- Requires DB-GPT environment; cannot be used standalone
- Fixed report template limits customization
- Performance on very large files is undocumented
- No evidence of unit tests or validation on non-standard platforms
How do you install this skill?
Since this skill is part of the DB-GPT repository, install DB-GPT first. Use the one-line installer: curl -fsSL https://raw.githubusercontent.com/eosphoros-ai/DB-GPT/main/scripts/install/install.sh | bash. Or via pip: pip install dbgpt-app. After installation, the skill files are located in skills/csv-data-analysis/ within the repo and are ready to use in DB-GPT.
How do you use this skill?
Upload a CSV or Excel file in DB-GPT and send a request like “Analyze CSV” or “Data analysis”. The skill automatically executes: 1) runs csv_analyzer.py to extract features; 2) calls html_interpreter with template_path and data parameters containing 9 placeholders (e.g., LANG, REPORT_TITLE, EXEC_SUMMARY). Example trigger prompt: “Please analyze this CSV file and generate a deep analysis report”.