skills.rest
superfhp
superfhpCOMMUNITY
Explore all skills by this author →
GitHub

report-writing

Generate self-contained HTML evaluation reports with ECharts visualizations from LLM benchmark CSV data.

Updated May 13, 2026
Search Tags:#data visualization#llm evaluation#model comparison#echarts#html report#csv parsing#finance benchmark
One-click install
npx skills add https://github.com/superfhp/lumi-agent-body --skill report-writing-superfhp
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: report-writing
Source: https://github.com/superfhp/lumi-agent-body/tree/main/skills/fin-llm-eval-report/report-writing
Command: npx skills add https://github.com/superfhp/lumi-agent-body --skill report-writing-superfhp

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? It turns raw LLM evaluation CSV files (knowledge QA, sentiment classification, research report generation) into a presentation-ready, single-file HTML report with interactive charts, eliminating manual data aggregation and report design work. ## Core Features & Use Cases - 7-Dimension Scoring: Aggregates three CSV types into seven finance capability dimensions with weighted composite scores and per-model rankings. - Interactive ECharts Reports: Produces radar charts, heatmaps, grouped bar charts, stability analysis, and root-cause sections, each with written interpretation. - Incomplete-Data Resilience: Degrades gracefully when datasets are partial, marking missing dimensions as N/A with remediation guidance instead of leaving blank sections. - Use Case: After benchmarking several LLMs on financial tasks, ask for an evaluation report and receive a timestamped HTML file served via a local web link, ready to present to stakeholders. ## Quick Start Generate a financial LLM evaluation report from the CSV data fetched by the fetch-data skill and give me the report access link.

Frequently Asked Questions about report-writing

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate an LLM evaluation report from CSV data?▼

Provide the three evaluation CSV files (FullReport, NewsReport, ResearchReport) via the fetch-data skill, then request a report. The skill parses scores, computes 7 weighted dimensions, and outputs a self-contained HTML file with ECharts visualizations.

What CSV formats are supported for model evaluation reports?▼

Three types are supported: knowledge QA with model/domain/difficulty/Accuracy columns, sentiment classification with model/difficulty/Accuracy, and research report generation with factuality_score and recall_score. Chinese column names are also recognized as fallbacks.

Can the report be generated when some evaluation data is missing?▼

Yes. Missing dimensions are shown as N/A with explanations and remediation steps, composite rankings degrade to computable subsets with disclaimers, and every section still contains at least one readable interpretation block.

How is the composite score calculated across evaluation dimensions?▼

Seven dimensions are weighted: research report ability 20%, four knowledge dimensions at 15% each, and sentiment classification plus quantitative analysis at 10% each. Research report scores combine factuality at 60% and recall at 40%.

Why are some models excluded from the overall ranking?▼

Models missing any of the three evaluation types cannot receive a complete composite score and are marked as pending completion. Models with abnormal latency over 1000 seconds or all-zero scores are flagged and excluded separately.

Related Skills

Explore More Data & Analytics Skills→
githubgithub
38.5k

eval-driven-dev

Builds evaluation pipelines for Python LLM applications using pixie-qa instrumentation and scoring.

Official
Advanced
KILWA73KILWA73

theme-factory

Select a theme from theme-showcase.pdf and apply it to slides, reports, docs, or HTML pages.

Community
Intermediate
erwinv2k-TKGerwinv2k-TKG

observability

Analyzes logs, metrics, and traces across observability backends using statistics-first methodology.

Community
Intermediate
inbharataiinbharatai
22

dashboard-builder

Build interactive data dashboards and reports with Streamlit and Dash.

Community
Intermediate