report-analyzer

Report Analyzer

Analyzes large reports (PDF, PPTX) and produces a structured summary with key data and insights.

Target audience: product managers, marketers, analysts, finance professionals, C-level executives.

---

Language Detection

Detect the language from the user's message. Use that language for all output — the summary document, clarifying questions, and responses. If the user writes in Russian, respond and generate the output in Russian. If in English — in English.

---

Input

• A PDF or PPTX report file
• The user specifies the file name or path in their message, or uploads the file directly

---

Output

A structured summary document (up to 1.5 pages) containing:

• Report metadata
• Executive Summary
• Key figures and data
• Key insights
• Report structure

---

Instructions

Step 1 — Locate the input file

The user specifies the report file name in their message. Locate the file using this order:

1. Check the current working directory and subdirectories (e.g. input/ folder or project root)
2. If the user provided a full path — use it directly

File search:

find . -iname "*.pdf" -o -iname "*.pptx" 2>/dev/null

If the file is not found — ask the user:

"Please specify the exact file name or path to the report, or upload the file directly. Supported formats: PDF, PPTX."

If the format is not supported:

"This skill works with PDF and PPTX files. Please provide a file in one of these formats."

---

Step 2 — Ask clarifying questions

Ask the user the following questions before proceeding:

Question 1: "What is the analysis focus?"

• General overview
• Numbers and data
• Strategic takeaways
• Everything combined

Question 2: "What output file format do you prefer?"

• .md
• .pdf
• .docx

Do not create the output file until both questions are answered.

---

Step 3 — Read and process the file

For PDF:

import pdfplumber

with pdfplumber.open("path/to/file.pdf") as pdf:
    text = ""
    for page in pdf.pages:
        text += page.extract_text() or ""
        # Extract tables
        tables = page.extract_tables()