JobSentinel

AI-powered job scam detection and validation platform.

Protect yourself from fake listings, ghost jobs, and recruitment fraud before you apply.

Website | API Docs | Browser Extension

---

Why JobSentinel?

Job scams cost victims $2 billion+ annually (FTC, 2024). Platform moderation catches obvious fakes, but sophisticated scams slip through — MLM recruitment disguised as corporate roles, ghost jobs that waste months of your time, data harvesting posts that steal your identity.

JobSentinel catches what platforms miss.

Quick Start

pip install -e .

# Analyze a suspicious posting
sentinel analyze "We're hiring! No experience needed. Earn $5000/week guaranteed. Send $50 registration fee to start."

# Analyze from a URL
sentinel analyze "https://linkedin.com/jobs/view/1234567890"

# Batch analyze from file
sentinel analyze --file jobs.json

# Validate a company
sentinel validate "Google"

# Report a scam (improves detection)
sentinel report "https://linkedin.com/jobs/view/123" --reason "Asked for SSN before interview"

# View detection statistics
sentinel stats

Install Options

pip install -e ".[full]"    # Everything: AI + API server + web scraping
pip install -e ".[ai]"      # AI analysis (anthropic)
pip install -e ".[api]"     # API server (fastapi + uvicorn)
pip install -e ".[web]"     # Web scraping (httpx + beautifulsoup4)

What It Detects

Red Flags

Signal	Example
Upfront payment	"Pay $99 registration fee to start"
Personal info harvesting	SSN, bank account before interview
Guaranteed income	"$5,000/week guaranteed"
Suspicious contact	Corporate role but gmail/yahoo
Crypto payment	"Paid in Bitcoin/crypto"
No company presence	No verifiable business presence
Interview bypass	"No interview needed, start immediately"
MLM / pyramid	"Build your team", "unlimited earning potential"
Reshipping scam	"Receive and forward packages"

Warnings

• Salary significantly above market rate
• Extremely vague job descriptions
• No qualifications listed
• Urgency / scarcity language
• Recently created company profiles
• Low-connection recruiters

Ghost Jobs

• Stale postings (30+ days, no activity)
• Serial reposts (same role monthly)

Risk Levels

Score	Level	Meaning
0.0-0.2	Safe	Proceed with confidence
0.2-0.4	Low	Likely legitimate
0.4-0.6	Suspicious	Review before applying
0.6-0.8	High	Strong scam indicators
0.8-1.0	Scam	Almost certainly fraudulent

Company Validation

Cross-reference companies against multiple sources:

• Known employers database — 200+ verified major employers
• Domain WHOIS — Check domain age (new domains = higher risk)
• LinkedIn company page — Verify follower count, employee count
• Result caching — 7-day TTL, --refresh to force re-check

Browser Extension (Chrome)

Analyze job postings directly in your browser:

# Load the extension in Chrome developer mode
# Navigate to: chrome://extensions > Load unpacked > sentinel/web/extension/

# Start the API server
sentinel serve --port 8080

The extension adds a color-coded risk badge next to each job title with a detailed breakdown popup.

REST API

pip install -e ".[api]"
sentinel serve --port 8080
# API docs at http://localhost:8080/docs

POST /api/analyze

curl -X POST http://localhost:8080/api/analyze \
  -H "Content-Type: application/json" \
  -d '{"text": "Earn $5000/week guaranteed! Send $50 fee to start.", "title": "Data Entry", "company": ""}'

{
  "scam_score": 0.95,
  "risk_level": "scam",
  "risk_label": "Almost Certainly Scam",
  "red_flags": [
    {"name": "upfront_payment", "detail": "Registration/training fee required"},
    {"name": "guaranteed_income", "detail": "Unrealistic income guarantee"}
  ],
  "warnings": [
    {"name": "vague_description", "detail": "No specific duties listed"}
  ],
  "signal_count": 7,
  "analysis_time_ms": 4.2
}

POST /api/report

curl -X POST http://localhost:8080/api/report \
  -H "Content-Type: application/json" \
  -d '{"url": "https://linkedin.com/jobs/view/123", "is_scam": true, "reason": "Asked for payment"}'

GET /api/patterns

curl http://localhost:8080/api/patterns?category=red_flag

GET /api/stats | GET /api/health

Detection statistics and service health check.

Configuration

# ~/.config/sentinel/config.toml

db_path = "~/.sentinel/sentinel.db"
ai_enabled = true
ai_model = "claude-haiku-4-5"
ai_model_deep = "claude-sonnet-4-6"
rate_limit_rpm = 60
cors_origins = ["http://localhost:3000"]
log_level = "INFO"

Development

pip install -e ".[dev,full]"

# Run tests
python -m pytest tests/ -v

# Lint
ruff check sentinel/

Security

• Input validation on all API fields (length limits, URL format)
• SQL injection prevention (parameterized queries throughout)
• Command injection prevention (domain validation before subprocess calls)
• HTML sanitization (script/style/event handler stripping)
• See tests/test_security.py for the full security test suite

Dependencies

Dependency	Purpose	Required?
Python 3.12+	Runtime	Yes
click	CLI framework	Yes
anthropic	AI analysis	Optional
fastapi + uvicorn	API server	Optional
httpx	URL fetching	Optional
beautifulsoup4	HTML parsing	Optional

Core detection runs on Python stdlib only — no pip packages required for the detection engine.

Contributing

1. Fork the repo
2. Create a feature branch
3. Write tests for your changes
4. Run the test suite and linter
5. Submit a PR

License

MIT