What problem does it solve?
Gathering Amazon product reviews at scale usually requires paid scraper APIs, browser automation, or Amazon credentials. This Skill collects written Amazon US reviews through Woot's public review endpoint with no login or API key, while preserving every attempt and error log so the resulting dataset remains auditable.
Core Features & Use Cases
- Batch Review Acquisition: Reads a TSV of ASINs with basic/full/max collection modes, runs up to two workers with retries and backoff, and reconciles partial runs by unioning all valid attempts with exact content dedupe.
- Auditable Evidence Trail: Saves every attempt JSON plus stdout/stderr logs, detects partial runs even when exit codes are zero, and writes an acquisition manifest with per-target status (complete, partial, failed, no reviews).
- Normalized JSONL Export: Merges duplicate reviews across ASIN variants, parses human-readable dates, builds stable SHA-256 content keys, and outputs analysis-ready JSONL with star ratings, verified-purchase flags, Vine flags, media URLs, and keyword hit labels.
- Use Case: A VOC analyst needs low-star complaint data for five competitor ASINs. They list the ASINs in a TSV, run the batch script, and receive a deduplicated JSONL file plus a manifest showing exactly which requests succeeded or failed.
Quick Start
Ask the agent to collect Amazon reviews for a list of ASINs using this skill, specifying basic or full mode, and export the normalized results to a JSONL file in the project's data/raw folder.
