ワンクリックで
report-reference
Quality standards and guidelines for generating analysis reports across all report-producing skills
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Quality standards and guidelines for generating analysis reports across all report-producing skills
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Generate a qualitative interview guide as an EDSL Survey with QuestionInterview, based on Geiecke & Jaravel (2026) methodology
Publish a completed study to GitHub - creates a repo with the analysis report as README.md, all charts, data, and a reproducibility footer linking to Expected Parrot
Generate a comprehensive, self-contained analysis report for a conjoint (choice-based) study with executive summary, methodology, part-worth utilities, segment analysis, and limitations
Design, execute, and analyze conjoint analysis studies - guides through attribute/level definition, experimental design, EDSL survey generation, and part-worth utility estimation
Write a detailed referee report for an academic paper in economics or a related discipline
Create AgentLists from web searches, descriptions, local files, or programmatic generation
| name | report-reference |
| description | Quality standards and guidelines for generating analysis reports across all report-producing skills |
| allowed-tools | Read |
| user-invocable | false |
This reference skill codifies quality standards for all report-generating skills in the EDSL Research plugin. Any skill that produces a report.md / report.html should read this file first and follow these guidelines.
Every analysis report MUST include these sections, in order:
The "Key Findings" section must contain 4-6 paragraphs of substantive prose written by Claude. Each paragraph should:
Never use placeholder text like "[Add key findings based on the analysis]" or "[Insert findings here]". Claude must write the actual findings based on the data.
When a survey uses Jinja2 scenario templates (e.g., {{ task_1_opt_a }}):
If the same question template is used for N tasks (common in conjoint studies), do NOT list it N times. Instead:
When displaying agent/respondent personas:
segment trait as the display label/headersegment trait value as the primary label for each agentpersona trait text under each segment labelAll images referenced in report.md MUST be in the same directory as the report:
Copy any images generated outside the output directory into the output directory before referencing them.
Report the analytical N (number of independent observations/choices), not just the CSV row count:
Convert report.md to report.html using:
pandoc report.md -o report.html --css=<css_path> --standalone
Glob("**/assets/report.css")--metadata title= (the markdown # heading serves as the title)Use these colors consistently across all charts:
| Purpose | Color | Hex |
|---|---|---|
| Primary / positive values | Steel blue | #4C78A8 |
| Negative values / warnings | Coral red | #E45756 |
| Secondary | Teal | #72B7B2 |
| Accent | Orange | #F58518 |
| Neutral | Gray | #BAB0AC |
Always use relative links for files generated alongside the report:
[results.csv](results.csv) not [results.csv](./analysis_1/results.csv)[utilities.json](utilities.json) not an absolute pathInclude a generation timestamp at the top of the report:
*Generated: YYYY-MM-DD HH:MM:SS*