sn-da-non-spreadsheet-analysis

Document Analysis Skill — Word / PDF / PPT

End-to-end workflow for Word, PDF, and PPT document parsing. Each format has specific parsing pitfalls — follow the format-specific sub-skill exactly.

---

Workflow

Step 0 — Identify file type and input scope

import os

input_path = "/mnt/data/..."  # from user

# Detect single file vs directory (multi-file scenario)
if os.path.isdir(input_path):
    all_files = [
        os.path.join(input_path, f)
        for f in os.listdir(input_path)
        if f.lower().endswith(('.docx', '.doc', '.pdf', '.pptx', '.ppt'))
    ]
    print(f"Found {len(all_files)} documents: {all_files}")
else:
    all_files = [input_path]

# Route by extension
ext = os.path.splitext(all_files[0])[-1].lower()
print(f"File type: {ext}")

Critical rule: When input_path is a directory OR the user says "这些文件" / "所有文档", process every file and aggregate. Never stop at the first file.

---

Step 1 — Load sub-skill by format

Extension	Sub-skill to load
.docx / .doc	capability/word-analysis/SKILL.md
.pdf	capability/pdf-analysis/SKILL.md
.pptx / .ppt	capability/ppt-analysis/SKILL.md

read_file(path="<skills_root>/sn-da-non-spreadsheet-analysis/capability/<format>-analysis/SKILL.md")

Load only the sub-skill you need — do not load all three at once.

---

Step 2 — Parse and extract

Follow the sub-skill's extraction pattern. For all formats:

• Full scan: iterate all pages/slides/paragraphs — never stop early
• Table extraction: get every table, not just the first one
• Image/chart detection: if a page/slide yields no text, treat it as image-based and call caption.py

---

Step 3 — Answer with verification

After extracting data, verify before answering:

# For count/statistics questions: spot-check 3-5 items
sample = result_list[:3]
print(f"Sample check: {sample}")
print(f"Total count: {len(result_list)}")

# For numeric calculations: print intermediate values
print(f"Max={max_val}, Min={min_val}, Range={max_val - min_val}")

# For unit-sensitive answers: always include the unit
print(f"Answer: {value} {unit}")  # e.g., "475 千港元" not just "475"