Report Processor

Automatically parse PDF/TXT research reports to extract key viewpoints, data, investment advice, risks, and generate summaries.

Security Scan

The skill's code and instructions match its stated purpose (parsing PDFs/TXT, calling a local Ollama model, saving JSON outputs); there are no hidden network endpoints or unrelated credential requests, but it will write extracted report contents to your home directory so review before use.

Purpose & Capability

Name/description align with the implementation: the script extracts text (pdftotext for PDFs), calls an LLM (Ollama qwen2.5:14b) to extract fields, and writes JSON results. Required system components (Ollama and poppler/pdftotext) are appropriate for the stated functionality.

Instruction Scope

SKILL.md instructs installing Ollama and poppler and running the provided script; the script only reads the provided files, invokes local tools, and writes outputs. It does not read unrelated host files or contact remote endpoints itself. Note: outputs (including partial model responses) are saved to ~/.openclaw/workspace/data/reports/ and might contain sensitive report contents.

Install Mechanism

This is an instruction-only skill with no install spec. SKILL.md suggests using standard installers (ollama pull and package manager installs for poppler). No downloads from untrusted URLs or archive extraction are present.

Credentials

The skill does not request credentials or environment variables. The script sets OLLAMA_HOST for the subprocess to 127.0.0.1:11434 (local Ollama), which is consistent with running a local model. There are no unexplained secrets or unrelated env access.

Persistence & Privilege

The skill does not request always:true and does not alter other skills or system-wide configurations. It writes output files under the user's home directory (~/.openclaw/workspace/data/reports/), which is expected for storing processed results but is persistent on disk.

Assessment

This skill appears coherent with its description, but review these points before installing or running: (1) It expects a local Ollama server and the qwen2.5:14b model; ensure Ollama is running locally and you have the model (SKILL.md instructs how). (2) Install poppler/pdftotext to enable PDF parsing. (3) The script saves extracted data and portions of the model response to ~/.openclaw/workspace/data/reports/ — the saved files may contain sensitive financial data, so consider changing OUTPUT_DIR, restricting permissions, or encrypting/deleting outputs if needed. (4) The script sets OLLAMA_HOST to 127.0.0.1:11434 for its subprocess call; do not change that to a remote host unless you understand the privacy implications (pointing it to a remote endpoint could send report contents off-host). (5) As a best practice, inspect the script and run it on non-sensitive sample data first to confirm behavior.

Report Processor (研报处理器)

Automatically collect, parse, and extract key information from research reports. Supports PDF, TXT, and other formats.

Features

• Parse research reports (PDF/TXT)
• Extract key information (core观点、数据、投资建议、风险)
• Auto summary generation
• Knowledge base storage

Requirements

• Ollama with qwen2.5:14b model installed
• poppler-utils (for PDF parsing)

Installation

# Install Ollama and pull model
ollama pull qwen2.5:14b

# Install poppler (for PDF support)
# macOS
brew install poppler
# Linux
sudo apt install poppler-utils

Usage

# Process single file
python3 scripts/report_processor.py /path/to/report.pdf

# Batch process directory
python3 scripts/report_processor.py /path/to/reports/

Output

Results are saved to ~/.openclaw/workspace/data/reports/ in JSON format.

Configuration

Edit scripts/report_processor.py to customize:

• OLLAMA_MODEL: Model to use (default: qwen2.5:14b)
• OUTPUT_DIR: Output directory