ppt+proposal: 素材摄取改用 markitdown, 删自研 source_to_md

ppt/proposal 的"素材 → Markdown"逻辑此前各写一份 (source_to_md.py
内联 pypdf/python-docx/openpyxl), 改用微软 markitdown CLI 统一替换:
表格/标题/列表保留更好, 同时多覆盖 xlsx/url/html/csv 等格式。

- requirements.txt: 加 markitdown[pdf,docx,pptx,xlsx]
- skills/ppt/SKILL.md: 资源行改成 markitdown 说明
- skills/proposal/SKILL.md: 阶段零 32 行 Python 代码 → 4 行 CLI
- skills/ppt/scripts/source_to_md.py: 删除 (157 行)
- PROGRESS.md: scripts 列表同步

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

PROGRESS.md

@@ -22,7 +22,7 @@
 
 ## 已完成关键能力
 
-**Phase 1-3**(2026 早期):骨架 + skill 系统 + run_python。所有工具基目录是用户当前 cwd(不是 zcbot 仓库本身),agent 操作的是用户项目。`tools/fs.py` 的 `edit` 用 CoreCoder 风格唯一匹配。`tools/run_python.py` 过滤 `*API_KEY *TOKEN *SECRET *PASSWORD *PRIVATE_KEY` 环境变量。三个 skill 中 `ppt/` 最完整(v3:商务红硬约束 + apply_brand 品牌条 + Iconify 图标库 + scripts:fetch_icon / quality_check / source_to_md / render_icon)。
+**Phase 1-3**(2026 早期):骨架 + skill 系统 + run_python。所有工具基目录是用户当前 cwd(不是 zcbot 仓库本身),agent 操作的是用户项目。`tools/fs.py` 的 `edit` 用 CoreCoder 风格唯一匹配。`tools/run_python.py` 过滤 `*API_KEY *TOKEN *SECRET *PASSWORD *PRIVATE_KEY` 环境变量。三个 skill 中 `ppt/` 最完整(v3:商务红硬约束 + apply_brand 品牌条 + Iconify 图标库 + scripts:fetch_icon / quality_check / render_icon;素材摄取改用 markitdown CLI)。
 
 **Phase 4**(2026-05-06):
 - `core/probe.py` + `cli.py probe` —— basic_chat / parallel_tools / thinking_mode / long_context 四项探测

requirements.txt

@@ -7,3 +7,6 @@ rich>=13.7.0
 python-pptx>=0.6.21
 python-docx>=1.1.0
 matplotlib>=3.8.0
+
+# 素材摄取: PDF/DOCX/PPTX/XLSX/HTML/URL → Markdown (ppt 阶段零 + proposal 阶段零)
+markitdown[pdf,docx,pptx,xlsx]>=0.0.1

skills/ppt/SKILL.md

@@ -12,7 +12,7 @@ description: 生成 PowerPoint 演示文稿 (.pptx)。当用户要求做汇报 P
 - `references/layouts.md` —— 9 种版式的 python-pptx 起手代码 + 安全区/越界保护 + `apply_brand` 品牌条
 - `references/icons.md` —— 业务图标两层:Iconify (在线/本地缓存) / unicode 字形兜底
 - `assets/icons/` —— 本地图标缓存 (Iconify 拉过的图存这,见 `INDEX.md` 推荐清单)
-- `scripts/source_to_md.py` —— PDF/DOCX/PPTX/URL → 干净 Markdown
+- 素材摄取: 直接用 `markitdown` CLI (PDF/DOCX/PPTX/XLSX/HTML/URL → 干净 Markdown)
 - `scripts/fetch_icon.py` —— 从 Iconify CDN 拉 SVG/PNG (染主题色,缓存本地)
 - `scripts/render_icon.py` —— unicode 字形 → 透明 PNG (Iconify 没有时兜底)
 - `scripts/quality_check.py` —— 产物 .pptx 验收 (越界 / 文本溢出 / 颜色一致)
@@ -94,7 +94,7 @@ python <skill_dir>/scripts/quality_check.py <task_dir>/<output.pptx> --spec <tas
 
 ```
 <task_dir>/
-├── source.md            # source_to_md.py 转出的素材
+├── source.md            # markitdown 转出的素材
 ├── spec_lock.md         # 八条对齐落定
 ├── slides/
 │   └── chart_p3.png     # 各页用到的图片素材

skills/ppt/scripts/source_to_md.py

@@ -1,157 +0,0 @@
-"""source_to_md.py: 把素材转成干净 Markdown,作为后续策略阶段的输入。
-
-用法:
-    python source_to_md.py <input>           # 自动按扩展名识别
-    python source_to_md.py <url>             # http/https 走 web 抓
-    python source_to_md.py file.pdf -o source.md
-
-支持:
-    .pdf  → pypdf 提取文本
-    .docx → python-docx 段落
-    .pptx → python-pptx 提取每页文字
-    .txt/.md → 直读
-    URL   → requests + 简易 HTML 剥离
-
-设计原则:模型在策略阶段只看 Markdown,不读二进制 / 不爬复杂排版。
-"""
-from __future__ import annotations
-
-import argparse
-import re
-import sys
-from pathlib import Path
-from urllib.parse import urlparse
-
-
-def from_pdf(path: Path) -> str:
-    try:
-        from pypdf import PdfReader
-    except ImportError:
-        return "[error] pip install pypdf"
-    reader = PdfReader(str(path))
-    parts = [f"# {path.stem}\n"]
-    for i, page in enumerate(reader.pages, 1):
-        text = (page.extract_text() or "").strip()
-        if text:
-            parts.append(f"\n## Page {i}\n\n{text}\n")
-    return "\n".join(parts)
-
-
-def from_docx(path: Path) -> str:
-    try:
-        from docx import Document
-    except ImportError:
-        return "[error] pip install python-docx"
-    doc = Document(str(path))
-    parts = [f"# {path.stem}\n"]
-    for para in doc.paragraphs:
-        text = para.text.strip()
-        if not text:
-            continue
-        style = (para.style.name or "").lower() if para.style else ""
-        if "heading 1" in style:
-            parts.append(f"\n## {text}\n")
-        elif "heading 2" in style:
-            parts.append(f"\n### {text}\n")
-        elif "heading 3" in style:
-            parts.append(f"\n#### {text}\n")
-        else:
-            parts.append(f"\n{text}\n")
-    return "".join(parts)
-
-
-def from_pptx(path: Path) -> str:
-    try:
-        from pptx import Presentation
-    except ImportError:
-        return "[error] pip install python-pptx"
-    prs = Presentation(str(path))
-    parts = [f"# {path.stem}\n"]
-    for i, slide in enumerate(prs.slides, 1):
-        parts.append(f"\n## Slide {i}\n")
-        for shape in slide.shapes:
-            if shape.has_text_frame:
-                txt = shape.text_frame.text.strip()
-                if txt:
-                    parts.append(f"\n{txt}\n")
-    return "".join(parts)
-
-
-def from_text(path: Path) -> str:
-    return path.read_text(encoding="utf-8", errors="replace")
-
-
-_TAG_RE = re.compile(r"<[^>]+>")
-_WS_RE = re.compile(r"\n{3,}")
-
-
-def from_url(url: str) -> str:
-    try:
-        import requests
-    except ImportError:
-        return "[error] pip install requests"
-    r = requests.get(url, timeout=30, headers={
-        "User-Agent": "Mozilla/5.0 (compatible; ppt-source-to-md/1.0)"
-    })
-    r.raise_for_status()
-    html = r.text
-
-    # 极简剥离:script/style 删,标签去除
-    html = re.sub(r"<script[\s\S]*?</script>", "", html, flags=re.I)
-    html = re.sub(r"<style[\s\S]*?</style>", "", html, flags=re.I)
-
-    title_m = re.search(r"<title[^>]*>([^<]+)</title>", html, re.I)
-    title = title_m.group(1).strip() if title_m else url
-
-    # 块级标签转换行
-    html = re.sub(r"</?(p|div|br|li|h[1-6]|tr)[^>]*>", "\n", html, flags=re.I)
-    text = _TAG_RE.sub("", html)
-    text = re.sub(r"&nbsp;", " ", text)
-    text = re.sub(r"&amp;", "&", text)
-    text = re.sub(r"&lt;", "<", text)
-    text = re.sub(r"&gt;", ">", text)
-    text = re.sub(r"&quot;", '"', text)
-    text = "\n".join(line.strip() for line in text.splitlines())
-    text = _WS_RE.sub("\n\n", text).strip()
-
-    return f"# {title}\n\nSource: {url}\n\n{text}\n"
-
-
-def dispatch(src: str) -> str:
-    parsed = urlparse(src)
-    if parsed.scheme in ("http", "https"):
-        return from_url(src)
-
-    path = Path(src)
-    if not path.exists():
-        return f"[error] not found: {src}"
-
-    ext = path.suffix.lower()
-    if ext == ".pdf":
-        return from_pdf(path)
-    if ext == ".docx":
-        return from_docx(path)
-    if ext == ".pptx":
-        return from_pptx(path)
-    if ext in (".txt", ".md"):
-        return from_text(path)
-    return f"[error] unsupported extension: {ext}"
-
-
-def main():
-    ap = argparse.ArgumentParser()
-    ap.add_argument("src", help="文件路径或 http(s) URL")
-    ap.add_argument("-o", "--output", type=Path, default=None,
-                    help="写到文件;默认打印到 stdout")
-    args = ap.parse_args()
-
-    md = dispatch(args.src)
-    if args.output:
-        args.output.write_text(md, encoding="utf-8")
-        print(f"[ok] {args.output}  ({len(md)} chars)")
-    else:
-        sys.stdout.write(md)
-
-
-if __name__ == "__main__":
-    main()

skills/proposal/SKILL.md

@@ -21,29 +21,15 @@ description: 撰写中国科研项目申报书 / 课题任务书 (国家重点
 - `<skill_dir>/scripts/word_count.py` —— 章节字数 vs 预算
 - `<skill_dir>/scripts/quality_check.py` —— 结构完整性 / 假大空话术 / 占位符未替换 / 指南覆盖度 (--spec 选项)
 
-## 阶段零: 摄取素材 (有 PDF/DOCX 时才走)
+## 阶段零: 摄取素材 (有 PDF/DOCX/XLSX/URL 时才走)
 
-用户给指南 PDF / 团队介绍 DOCX / 预算 XLSX → 先转成 `<task_dir>/source/<name>.md`,后续阶段一才能读。用 `run_python` 即可,不需要新工具:
+用户给指南 PDF / 团队介绍 DOCX / 预算 XLSX / 政策网页 URL → 先转成 `<task_dir>/source/<name>.md`,后续阶段一才能读。统一用 `markitdown` CLI,表格 / 列表 / 标题层级会自动保留:
 
-```python
+```bash
-# PDF (指南文件)
+markitdown <path>/guide.pdf       -o <task_dir>/source/guide.md
-from pypdf import PdfReader
+markitdown <path>/team.docx       -o <task_dir>/source/team.md
-text = "\n\n".join(p.extract_text() or "" for p in PdfReader(pdf_path).pages)
+markitdown <path>/budget.xlsx     -o <task_dir>/source/budget.md
-Path("<task_dir>/source/guide.md").write_text(text, encoding="utf-8")
+markitdown https://example.com/x  -o <task_dir>/source/policy.md
-
-# DOCX (团队/前期成果)
-from docx import Document
-doc = Document(docx_path)
-md = "\n".join(p.text for p in doc.paragraphs if p.text.strip())
-# 表格
-for t in doc.tables:
-    for row in t.rows: md += "\n| " + " | ".join(c.text.strip() for c in row.cells) + " |"
-
-# XLSX (预算)
-from openpyxl import load_workbook
-wb = load_workbook(xlsx_path)
-for ws in wb.worksheets:
-    for row in ws.iter_rows(values_only=True): print(row)
 ```
 
 转完后 spec_lock 阶段直接 `read <task_dir>/source/*.md` 拿事实,不要凭印象写。