ChatGPT 内部Skill系统架构解析

oai
├── .bash_logout
├── .bashrc
├── .chromium
│   ├── Default
│   │   └── Preferences
│   └── Local State
├── .config
│   └── openbox
│       └── rc.xml
├── .ipython
│   └── profile_default
│       ├── db
│       ├── history.sqlite
│       ├── log
│       ├── pid
│       ├── security
│       └── startup
│           └── README
├── .nssdbp
├── .pki
│   └── nssdb
│       ├── cert9.db
│       ├── key4.db
│       └── pkcs11.txt
├── .profile
├── redirect.html
├── share
│   └── slides
│       ├── create_montage.py
│       ├── ensure_raster_image.py
│       ├── pptxgenjs_helpers
│       │   ├── code.js
│       │   ├── image.js
│       │   ├── index.js
│       │   ├── latex.js
│       │   ├── layout_builders.js
│       │   ├── layout.js
│       │   ├── svg.js
│       │   └── util.js
│       ├── render_slides.py
│       └── slides_test.py
├── skills
│   ├── docs
│   │   ├── render_docx.py
│   │   └── skill.md
│   ├── pdfs
│   │   └── skill.md
│   └── spreadsheets
│       ├── artifact_tool_spreadsheet_formulas.md
│       ├── artifact_tool_spreadsheets_api.md
│       ├── examples
│       │   ├── create_basic_spreadsheet.py
│       │   ├── create_spreadsheet_with_styling.py
│       │   ├── features
│       │   │   ├── change_existing_charts.py
│       │   │   ├── cite_cells.py
│       │   │   ├── create_area_chart.py
│       │   │   ├── create_bar_chart.py
│       │   │   ├── create_doughnut_chart.py
│       │   │   ├── create_line_chart.py
│       │   │   ├── create_pie_chart.py
│       │   │   ├── create_tables.py
│       │   │   ├── set_cell_borders.py
│       │   │   ├── set_cell_fills.py
│       │   │   ├── set_cell_width_height.py
│       │   │   ├── set_conditional_formatting.py
│       │   │   ├── set_font_styles.py
│       │   │   ├── set_merge_cells.py
│       │   │   ├── set_number_formats.py
│       │   │   ├── set_text_alignment.py
│       │   │   └── set_wrap_text_styles.py
│       │   ├── read_existing_spreadsheet.py
│       │   └── styling_spreadsheet.py
│       ├── skill.md
│       └── spreadsheet.md
└── .wgetrc

23 directories, 55 files

skills/
├── spreadsheets/
│   ├── skill.md                    # Skill定义文档（136行）
│   ├── artifact_tool_spreadsheets_api.md  # API 文档（1188行）
│   └── examples/
│       └── create_basic_spreadsheet.py
├── pdfs/
│   └── skill.md                    # PDF Skill定义
├── docs/
│   └── skill.md                    # DOCX Skill定义
└── slides/
    └── ...

# Spreadsheet Skill (Create • Edit • Analyze • Visualize)

Use this skill when you need to work with spreadsheets (.xlsx, .csv, .tsv) to do any of the following:
- Create a new workbook/sheet with proper formulas, cell/number formatting, and structured layout
- Read or analyze tabular data (filter, aggregate, pivot, compute metrics) directly in a sheet
- Modify an existing workbook without breaking existing formulas or references
- Visualize data with in-sheet charts/tables and sensible formatting
- Recalculate/evaluate formulas to update results after changes

from artifact_tool import SpreadsheetArtifact

# 创建新的电子表格
spreadsheet = SpreadsheetArtifact("AwesomeCompany")

# 创建工作表
employees_sheet = spreadsheet.create_sheet("Employees")

# 设置单元格值
employees_sheet.range("A1:D1").set_values([["Title", "Name", "Address", "Score"]])

# 添加公式
employees_sheet.cell("D6").formula = "=SUM(D2:D4)"

# 重新计算公式
employees_sheet.recalculate()

# 渲染为图像进行检查
render_path = spreadsheet.render()

# 导出为 XLSX
export_path = spreadsheet.export()

from_source_file() → 编辑操作 → recalculate() → render() → export()

### Check your work
- Before providing your completed spreadsheet to the user, check to ensure it is accurate and without errors.
- Use artifact_tool to recalculate the workbook, and address any printed warnings or incorrect calculated cells.
- After recalculating to ensure calculated values are cached, render the spreadsheet with artifact_tool
  and view the images to check for style, formatting, and correctness

### Finance-specific color conventions:
When building new financial models where no user-provided spreadsheet or formatting instructions override these conventions, use:
- **Blue text (RGB: 0,0,255)**: Hardcoded inputs, and numbers users will change for scenarios
- **Black text (RGB: 0,0,0)**: ALL formulas and calculations
- **Green text (RGB: 0,128,0)**: Links pulling from other worksheets within same workbook
- **Red text (RGB: 255,0,0)**: External links to other files
- **Yellow background (RGB: 255,255,0)**: Key assumptions needing attention or cells that need to be updated

### Additional requirements for investment banking
- Total calculations should sum a range of cells directly above them.
- Hide gridlines. Add horizontal borders above total calculations, spanning the full range of relevant columns.
- Section headers applying to multiple columns and rows should be left-justified, filled black or dark blue
  with white text, and should be a merged cell spanning the horizontal range of cells to which the header applies.
- Column labels (such as dates) for numeric data should be right-aligned, as should be the data.

### Ensure formulas are correct
- Spreadsheets must not contain added formula errors (#REF!, #DIV/0!, #VALUE!, #N/A, #NAME?)
- Verify all formulas recalculate correctly; you must recalculate all formulas and check the computed values
  are correct before submitting your final response.
- Guard against bugs in formulas:
  - Verify correct cell references
  - Check for off-by-one errors
  - Test by varying inputs with edge cases
  - Verify no unintended circular references

### Do not use dynamic array functions
- DO NOT use dynamic array formulas like FILTER, XLOOKUP, SORT, SEQUENCE
- These are not currently supported in openpyxl or artifact_tool in a way that is compatible with
  modern spreadsheet applications, and using them could prevent the user from being able to use the spreadsheet.

def rasterize(pptx_path: str, out_dir: str, dpi: int) -> Sequence[str]:
    """Rasterise PPTX to PNG files placed in out_dir and return the image paths."""

    # 步骤 1: PPTX → PDF
    pdf_path = convert_to_pdf(pptx_path, user_profile, convert_tmp_dir, stem)

    # 步骤 2: PDF → PNG（使用 pdf2image）
    paths_raw = convert_from_path(
        pdf_path,
        dpi=dpi,
        fmt="png",
        thread_count=8,  # 并发处理
        output_folder=out_dir,
        paths_only=True,
        output_file="slide",
    )

    return final_paths

def calc_dpi_via_ooxml(input_path: str, max_w_px: int, max_h_px: int) -> int:
    """Calculate DPI from OOXML `ppt/presentation.xml` slide size (cx/cy in EMUs)."""
    with ZipFile(input_path, "r") as zf:
        xml = zf.read("ppt/presentation.xml")
    root = ET.fromstring(xml)

    # 从 OOXML 元数据读取幻灯片尺寸（单位：EMU）
    cx = int(sld_sz.get("cx") or 0)
    cy = int(sld_sz.get("cy") or 0)

    # 转换为英寸
    width_in = cx / EMU_PER_INCH
    height_in = cy / EMU_PER_INCH

    # 计算最佳 DPI
    return round(min(max_w_px / width_in, max_h_px / height_in))

def convert_to_pdf(pptx_path: str, user_profile: str, convert_tmp_dir: str, stem: str) -> str:
    # 尝试直接转换：PPTX → PDF
    run_cmd_no_check(cmd_pdf)
    if exists(pdf_path):
        return pdf_path

    # 回退方案：PPTX → ODP → PDF
    # 理由：通过 ODF 序列化器规范化 PPTX 特定结构，
    # 通常可以绕过 Impress PDF 导出在问题文档上的问题
    run_cmd_no_check(cmd_odp)
    if exists(odp_path):
        run_cmd_no_check(cmd_odp_pdf)
        if exists(pdf_path):
            return pdf_path

    return ""

生成 → 渲染 → 检查 → 修复 → 重新渲染 → 再次检查 → 交付