Amazon Review Checker

Amazon review authenticity analyzer. Detect fake reviews, suspicious patterns, and rating manipulation. Includes time clustering detection, content similarit...

Security Scan

The skill's code and runtime instructions align with its stated purpose (local review parsing, analysis, and HTML report generation); nothing requests unrelated credentials or system-wide access.

Purpose & Capability

The bundled Python scripts (parser, analyzer, HTML report generator) implement review parsing, multiple detection dimensions, and report output that match the skill description. There are no unexpected credentials, binaries, or unrelated dependencies declared.

Instruction Scope

SKILL.md keeps runtime instructions scoped to parsing pasted or JSON reviews and running the local scripts. It does recommend an npx-based installer command (which would fetch a package from the network) and hints at additional 'Account Profile Analysis' if more reviewer data is provided — those are optional and not implemented in the included code. The skill will process any review text you paste, which may include reviewer names or other PII provided by you.

Install Mechanism

There is no formal install spec in the package (instruction-only), so nothing is automatically written to disk by the registry. The README suggests using an npx command to add the skill from an external package (npx will fetch code from npm). The included code itself does not perform network calls; only the generated HTML references Chart.js via a CDN when viewed in a browser.

Credentials

The skill requires no environment variables, credentials, or config paths. The Python code does not access environment variables or external services. The only external resource is a Chart.js CDN link embedded in the generated HTML (client-side).

Persistence & Privilege

The skill is not marked always:true and does not request persistent system privileges. It does not modify other skills or system-wide agent configuration. It generates a local HTML file if instructed, which is normal for reporting tools.

Assessment

This package appears coherent for analyzing Amazon reviews, but take these precautions before running it: 1) Inspect the included Python files locally (they are bundled) before execution — they appear to be plain parsing/analysis code with no network calls. 2) Run scripts in a controlled environment (virtualenv or sandbox) especially if you plan to supply large or unknown inputs. 3) Be careful pasting real review text that contains reviewer names, emails or other PII — the tool will process whatever you provide. 4) If you choose to run the npx command from SKILL.md, remember npx will fetch code from npm (network) — verify the package/source and prefer running the bundled scripts directly if you trust the bundled files. 5) The HTML report loads Chart.js from a CDN when opened in a browser; if you need offline or air-gapped operation, change or remove that reference. 6) If you want higher assurance, run static scans on the code or execute it in a restricted environment and observe network activity before providing any sensitive data.

Review authenticity analyzer — detect fake reviews, suspicious patterns, and rating manipulation.

Installation

npx skills add nexscope-ai/eCommerce-Skills --skill amazon-review-checker -g

Features

• Authenticity Score — 0-100 comprehensive rating
• Suspicious Pattern Detection — Time clustering, content similarity, rating anomalies
• Fake Review Flagging — Mark high-risk reviews with explanations
• Progressive Analysis — More data = deeper insights

Progressive Analysis Levels

Level	Required Data	Unlocked Analysis
L1 Basic	Review content	Similarity, length, keywords
L2 Advanced	+ Review date	Time clustering detection
L3 Deep	+ Star rating	Rating distribution analysis
L4 Complete	+ VP status	Verified purchase validation

Detection Dimensions

Dimension	Weight	Method
Time Clustering	25%	Sliding window + burst detection
Content Similarity	20%	N-gram + Jaccard similarity
Rating Distribution	20%	Chi-square test vs natural distribution
VP Ratio	15%	Compare to category benchmark
Review Length	5%	Entropy analysis
Suspicious Keywords	5%	Keyword pattern matching

Risk Levels

Score	Level	Description
70-100	✅ Low Risk	Reviews appear authentic
50-69	⚠️ Medium Risk	Some concerns found
30-49	🔴 High Risk	Multiple red flags
0-29	💀 Critical	Likely mass fake reviews

Usage

Method 1: Paste Reviews

Paste reviews directly in conversation:

Check these reviews:

5 stars - Great product! Works perfectly.
5 stars - Amazing! Best purchase ever.
1 star - Not as described.

Method 2: JSON Input

python3 scripts/analyzer.py '[
  {"content": "Great product!", "rating": 5, "date": "2024-01-15", "verified_purchase": true},
  {"content": "Amazing!", "rating": 5, "date": "2024-01-15", "verified_purchase": false}
]'

Method 3: Demo Mode

python3 scripts/analyzer.py --demo

Output Example

📊 Review Authenticity Report

ASIN: B08XXXXX
Reviews: 10
Analysis Level: L4

━━━━━━━━━━━━━━━━━━━━━━━━

Authenticity Score: 66/100 ⚠️

Medium Risk - Some concerns found.

━━━━━━━━━━━━━━━━━━━━━━━━

Detection Dimensions

🔴 Time Clustering: 70/100
   Max 6 reviews within 48h

✅ Content Similarity: 24/100
   Found 0 highly similar review groups

━━━━━━━━━━━━━━━━━━━━━━━━

High-Risk Reviews (Top 3)

1. Risk 75% - "Perfect!"
   Reason: Too short, non-VP, templated 5-star

🔍 Want more accurate analysis? Add:
• Reviewer info → Unlock "Account Profile Analysis"

Interaction Flow

User Input (any format)
        ↓
Smart field detection
        ↓
Analyze with available data
        ↓
Results + depth suggestions
        ↓
User continues or ends

---

Part of Nexscope AI — AI tools for e-commerce sellers.