windylam

Collect and locally process ride-sharing receipts from Gmail into structured data and SQLite for spending and behavior insights, ensuring privacy.

This skill appears to do exactly what it says: it uses the gog CLI to fetch ride receipts from a selected Gmail account, stores the raw email JSON/HTML locally, sends that raw payload to a Gateway model running on localhost for extraction, and loads the extracted records into a local SQLite DB and anonymized CSV. Before installing/run it: (1) ensure you have and trust a local OpenClaw Gateway instance (the skill refuses non-local hosts), (2) confirm you are comfortable with raw receipt HTML/JSON being written to data/ride-insights/emails.json and sent to the local model, (3) protect the OPENCLAW_GATEWAY_TOKEN and the ~/.openclaw/openclaw.json file, (4) review and delete emails.json if you do not want the raw receipts to persist, and (5) ensure the gog CLI is authenticated only for the account(s) you intend to process. If you need remote/external extraction or do not want raw emails written to disk, do not install or run this skill.

Type: OpenClaw Skill Name: windylamdatahive Version: 1.0.0 The skill bundle is a privacy-focused tool for analyzing ride-sharing receipts from Gmail. It includes scripts for fetching emails via the 'gog' CLI (fetch_emails_json.py), extracting data using a local LLM gateway (extract_rides_gateway.py), and generating an anonymized CSV for export (export_anonymized_rides_csv.py). The code contains explicit security safeguards, such as a strict check in extract_rides_gateway.py to ensure sensitive email data is only sent to local loopback addresses (localhost/127.0.0.1), and the SKILL.md instructions require the agent to obtain user confirmation before processing sensitive data.