中文 EN

Scan Report

Scan New Files

✓ Trusted — Risk Score 5/100
Last scan:2 days ago Rescan
5 /100
amazon-review-workbook
Collect Amazon product reviews via Chrome CDP and export a 14-column delivery-ready workbook with optional DeepLX translation and model-assisted semantic tagging.
Legitimate Amazon review scraper using Chrome CDP; all operations are documented, scoped, and appropriate for the stated purpose.
Skill Nameamazon-review-workbook
Duration49.5s
Enginepi
✓
Safe to install
No action needed. The skill is safe to use as documented.

Findings 5 items

Severity Finding Location
Low
No dependency version pinning
requirements.txt/package.json not present; requests library pinned to * in code imports. No known CVEs for used versions.
import requests
→ Add a requirements.txt with pinned versions if distributing as a package
 scripts/amazon_review_workbook.py:1
Info
Chrome CDP browser automation
Uses Chrome remote debugging protocol (port 9222) to scrape Amazon. All CDP eval expressions are reviewed; no data exfiltration beyond what is scraped from Amazon.
BrowserSession class
→ None; this is the documented mechanism
 scripts/amazon_review_workbook.py:657
Info
External network calls only to user-configured DeepLX
Network requests go to a URL configured by the user via DEEPLX_API_URL environment variable. No hardcoded external IPs.
requests.post(api_url, ...)
→ None; this is the documented translation feature
 scripts/deeplx_translate.py:95
Info
No sensitive path access
Script does not access ~/.ssh, ~/.aws, .env (other than its own optional .env for DeepLX config), or any credential stores outside its documented scope.
No os.environ iteration for secrets
→ None
 scripts/amazon_review_workbook.py:1
Info
No hidden functionality
All code paths are traceable to documented SKILL.md capabilities. CDP expressions are visible inline. No obfuscation, eval(), or base64 payloads.
browser.eval(expression)
→ None
 scripts/amazon_review_workbook.py:637
ResourceDeclaredInferredStatusEvidence
Filesystem READ READ ✓ Aligned Reads JSON/CSV/XLSX output files and local .env
Network READ READ ✓ Aligned Makes HTTP POST to user-configured DeepLX endpoint only
Shell WRITE WRITE ✓ Aligned Executes Python scripts via CLI; standard for script-based skills
Environment NONE READ ✓ Aligned Reads DEEPLX_API_URL/KEY from environment; does not harvest credentials
Skill Invoke NONE NONE — No skill invocation chains detected
Clipboard NONE NONE — No clipboard access
Browser READ READ ✓ Aligned Chrome CDP on localhost:9222; strictly scoped to scraping
Database WRITE WRITE ✓ Aligned SQLite cache for review data; contained in output directory
2 findings
🔗
Medium External URL 外部 URL
https://your-deeplx-host/translate
README.md:53
🔗
Medium External URL 外部 URL
http://127.0.0.1:
scripts/amazon_review_workbook.py:675

File Tree

12 files · 175.0 KB · 5210 lines
Python 5f · 4471L Markdown 6f · 735L YAML 1f · 4L
├─ ▾ 📁 agents
│ └─ 📋 openai.yaml YAML 4L · 352 B
├─ ▾ 📁 references
│ ├─ 📝 output-schema.md Markdown 81L · 2.4 KB
│ ├─ 📝 setup.md Markdown 104L · 3.1 KB
│ └─ 📝 tagging-guidelines.md Markdown 129L · 4.1 KB
├─ ▾ 📁 scripts
│ ├─ 🐍 amazon_review_workbook.py Python 2415L · 83.8 KB
│ ├─ 🐍 deeplx_translate.py Python 330L · 9.6 KB
│ ├─ 🐍 label_workflow.py Python 663L · 20.3 KB
│ ├─ 🐍 review_cache.py Python 465L · 15.7 KB
│ └─ 🐍 review_delivery_schema.py Python 598L · 18.6 KB
├─ 📝 LICENSE.zh-CN.md Markdown 23L · 1.1 KB
├─ 📝 README.md Markdown 240L · 6.4 KB
└─ 📝 SKILL.md Markdown 158L · 9.7 KB

Dependencies 5 items

PackageVersionSourceKnown VulnsNotes
requests * pip No No known CVE for current version; should be pinned for production
websocket-client * pip No Used for Chrome CDP WebSocket
pandas * pip No Workbook generation
openpyxl * pip No XLSX output
sqlite3 built-in stdlib No Local caching

Security Positives

✓ Comprehensive SKILL.md with explicit capability declarations
✓ Clean separation of concerns across 5 focused Python modules
✓ Chrome CDP access strictly scoped to localhost:9222
✓ No credential harvesting beyond optional DeepLX API key (used only for translation)
✓ No data exfiltration — all scraped data stays in local output directory
✓ SQLite cache contained within output directory, not system paths
✓ No subprocess or shell command injection vectors
✓ No base64/eval/atob obfuscation patterns
✓ No remote script download (curl|bash, wget|sh)
✓ No iteration through os.environ for sensitive keys
✓ No access to ~/.ssh, ~/.aws/credentials, or similar sensitive paths
✓ Well-structured code with type hints and error handling
✓ Explicit failure boundaries documented in SKILL.md