Test Report Generator

Skill: test-report-generator

Test Report Generator

Generate a formal hardware electromagnetic test report (检测报告) whose format matches the示范 file 生成报告示范.docx, driven by the data files in the working directory. The report is generic — it handles any measurement data type (RCS, 反射损耗, 复介电常数, 屏蔽效能, …), creating one test subsection per data file.

Generation is co-creative: ask the user for every variable, and default anything the user does not answer to the示范 report's own values (see references/defaults.md). The 结论 (conclusion) is NOT copied from the sample — it must be drafted from the actual data and charts.

How it works

The示范 document is used as a fixed "shell": its 封面, 说明, 目录, 1.试验概况 and 3/4/5 section headings are kept. scripts/report_generator.py fills the variable metadata and rebuilds the dynamic parts (2.试验内容和结果 per data file, 5.附件 data tables) and the conclusion.

Step 1 — Discover data files

Scan the working directory for CSV / TXT / Excel files. Preview each:

python .opencode/skills/test-report-generator/scripts/data_reader.py "<file_path>"

Auto-classify by filename/columns (RCS / 反射损耗 / 复介电常数 / 屏蔽效能 / 通用).

Ask the user: "发现以下数据文件，是否都用于本次报告？" (list them, with row/column counts). The user may exclude files.

Step 2 — Collect metadata (co-create, with defaults)

Ask the user for each item, showing the示范 default. The user may answer all at once, or answer "默认" / skip to accept the sample value.

If the user changes 客户/样品, also ask whether the narrative text (1.1 任务来源、1.2 被试品 等) should be updated accordingly.

Step 3 — Confirm data parsing per file (co-create)

For each data file, auto-detect and confirm with the user:

• X column (first numeric/frequency column)
• Y columns (remaining numeric columns)
• chart type (line for frequency sweeps, scatter for discrete points)
• statistics (mean/min/max; bandwidth/peak for 反射损耗; similarity for RCS)

Ask: "文件 X：X轴=列, Y轴=列1,列2…，图表=折线，统计=均值/最值。是否正确？"

Step 4 — Compute statistics & charts

python .opencode/skills/test-report-generator/scripts/statistics.py "<file>" "<x_col>" "<y_col>" "mean,min,max"
python .opencode/skills/test-report-generator/scripts/chart_generator.py "<file>" "<x_col>" "<y1,y2>" "line" "<title>" "<out.png>"

For RCS with comparable groups, also compute similarity and include it in the analysis text.

Step 5 — Draft analysis + conclusion (co-create)

• For each data file, draft a 2–3 sentence analysis from its stats/chart. Show the user and let them edit.
• Draft the conclusion (第4章) from ALL results + charts. This is not copied from the sample. Show it to the user for confirmation.

Step 6 — Assemble JSON and generate

Build report_data.json (schema below), then:

python .opencode/skills/test-report-generator/scripts/report_generator.py "<report_data.json>" "<output.docx>"

Save to <working_dir>/检测报告_<YYYYMMDD>_<HHMMSS>.docx. Tell the user the path.

JSON schema

{
  "report_no": "SXCL-20260322-1",
  "client": {"name": "浙江大学", "address": "...", "sample_name": "...", "sample_quantity": "1"},
  "timeline": {"receipt_date": "2026.3.21", "start_date": "2026.3.21", "end_date": "2026.3.22", "issue_date": "2026.3.23"},
  "meta": {"prepared_by": "杨欢", "reviewed_by": "陈波波", "approved_by": "李向洋",
           "organization": "陕西长岭电子科技有限责任公司", "address": "...", "phone": "..."},
  "equipment": "...",                 // optional; default = sample
  "test_location": "...",
  "environment": {"temperature": "...", "humidity": "...", "interference": "..."},
  "test_groups": [
    {"name": "反射损耗测试", "purpose": "...", "procedure": "...",
     "stats": {"headers": [...], "rows": [[...]]},
     "chart_path": "...", "chart_caption": "图1 ...", "analysis": "..."}
  ],
  "issues": "无。",
  "conclusion": ["段落1", "段落2"],
  "appendix": [{"name": "表1 ...", "headers": [...], "rows": [[...]]}]
}

Every field is optional; anything omitted keeps the sample default. test_groups = one per data file. conclusion is a list of paragraphs (drafted, not copied).

Processing rules

• Defaults: always from references/defaults.md. Never invent new defaults.
• Conclusion: always drafted from data + charts, never copied from the sample.
• Narrative (1.1–1.9, 3): default = sample text; only change when the user asks.
• Charts: use chart_generator.py (its SimHei font handling avoids missing CJK glyphs).
• Encoding: CSVs with UTF-8 BOM are handled by the scripts (chardet → UTF-8-SIG).
• Determinism: same JSON + same PNGs → same .docx.
• Report language: 中文.

Base directory: C:\Users\ganyangyu\Desktop\agent skill\.opencode\skills\test-report-generator