# JS Secrets Extraction

> Analyze JS bundles and source maps for hardcoded secrets, API keys, JWTs, and internal endpoints

- Skill: `uphiago/js-secrets-extraction` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add uphiago/js-secrets-extraction`
- Raw SKILL.md: https://api.skillmd.com/api/skills/uphiago/js-secrets-extraction/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Integrations & APIs
- License: MIT
- Author: uphiago (https://skillmd.com/u/uphiago)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/uphiago/js-secrets-extraction

---


# JS Bundle & Source Map Analysis -- Secret Extraction

## When to Use

- **ALWAYS** after initial web enumeration
- When you find modern SPA (React, Angular, Vue)
- When target uses Firebase, Supabase, Auth0
- Higher yield than directory scanning on many targets

## Why Analyze JS Bundles

Modern JavaScript bundles (Webpack, Vite, esbuild) often contain:
- Hardcoded API keys and tokens
- Internal API URLs
- Firebase, Auth0, Supabase configurations
- Environment variables (VITE_*, REACT_APP_*, NEXT_PUBLIC_*)
- Internal routes

## Bundle Download and Analysis

```bash
curl --max-time 30 --connect-timeout 10 -s "https://target.com" > index.html
grep -Eo 'src="[^"]*\.js"' index.html | cut -d'"' -f2 | while read js; do
  curl --max-time 30 --connect-timeout 10 -s "https://target.com$js" > "$(basename $js)"
done

# Search for secrets in bundles
grep -rEn "(apiKey|api_key|API_KEY|token|secret|password|clientId|client_id|auth0|firebase|supabase)[\"'\"]?[[:space:]]*[:=][[:space:]]*[\"''][^\"'']{8,}" *.js
```

## Source Map Reconstruction

```bash
curl --max-time 30 --connect-timeout 10 -sI "https://target.com/assets/index-abc123.js.map"
curl --max-time 30 --connect-timeout 10 -sI "https://target.com/static/js/main.12345.js.map"

# If HTTP 200, use for reconstruction:
# https://unminify.com
# https://source-map-visualization.netlify.app
```

**Real-world case**: Enterprise Angular SPA admin, 2 JS bundles (250KB each) exposed:
- Internal API URL (apiv3.empresa.com.br)
- Firebase API key (AIzaSy...2GXA)
- Encryption keys (AD5oDjsJaTJOzLe1Llj9mz)
- Cloudinary upload endpoint

## Port-Specific URL Analysis

Modern deployments often serve the main SPA on port 443 and admin/API on separate ports (8080, 8081, 8084). **Always check JS bundles on ALL discovered ports:**

```bash
# Check source maps on every open port
for port in 443 8080 8081 8084; do
  curl --max-time 30 --connect-timeout 10 -sI "https://target.com:$port/static/js/main.*.js.map" 2>/dev/null
  curl --max-time 30 --connect-timeout 10 -sI "https://target.com:$port/assets/index-*.js.map" 2>/dev/null
done
```

Source maps on administrative or alternate-port applications may expose a
different route and configuration set from the public SPA. Analyze each
authorized application independently.

## Admin Portal JS Analysis Pattern

When you find an admin portal on a separate port, the JS bundle often contains different secrets than the main site:

```python
base = "https://target.com:8080"  # Admin portal
js = requests.get(f"{base}/static/js/main.*.js").text

# 1. Extract ALL API URLs
api_urls = re.findall(r'https?://[^\"'[[:space:]]\\n,)>\\]]+', js)
# 2. Find base API URL (the backend this admin talks to)
# 3. Look for hardcoded credentials, API keys, auth patterns
# 4. Extract route paths for the admin app
routes = re.findall(r'[\"'](/[a-zA-Z0-9_/.-]*(?:admin|chat|bot|message|user|auth|login|token|config|setting|dashboard|hospital|pharmacy|drug|payment)[a-zA-Z0-9_/.-]*)[\"']', js, re.IGNORECASE)
```

## Source Map Content Analysis (1,200+ Files)

When source maps are available, analyze the `sourcesContent` array for hardcoded secrets:

```python
import json, re
data = json.loads(open("bundle.js.map").read())
all_source = " ".join(data.get("sourcesContent", []))

# Search for credentials in the original source
patterns = {
    "password": r'[\"']([^\"']*(?:password|passwd|pwd)[^\"']*)[\"']\s*[:=]\s*[\"']([^\"']+)[\"']',
    "token": r'[\"']([^\"']*(?:token|jwt|api_key|apikey|secret)[^\"']*)[\"']\s*[:=]\s*[\"']([^\"']+)[\"']',
}
for name, pat in patterns.items():
    matches = re.findall(pat, all_source, re.IGNORECASE)
    if matches:
        print(f"[{name}] {matches[:5]}")
```
- Cloudinary upload endpoint

## Secret Regex Patterns Catalog

```python
import re

patterns = {
    "Firebase API Key": r'apiKey:\s*[\"']([^\"']{30,})',
    "AWS Key": r'(?:AKIA|ASIA)[A-Z0-9]{16}',
    "Google API Key": r'AIza[0-9A-Za-z\\-_]{35}',
    "JWT": r'eyJ[A-Za-z0-9_\\-]{20,}\.[A-Za-z0-9_\\-]{20,}\.[A-Za-z0-9_\\-]{10,}',
    "Mercado Pago": r'APP_USR-[a-f0-9]{8,}',
    "Stripe": r'(?:sk_live|pk_live)_[A-Za-z0-9]{24,}',
    "Auth0 Domain": r'(?:domain|auth0_domain):\s*[\"']([^\"']+\.auth0\.com)',
    "Auth0 Client ID": r'(?:client_id|clientId|AUTH0_CLIENT_ID):\s*[\"']([^\"']{20,})',
    "Supabase URL": r'(?:supabaseUrl|SUPABASE_URL):\s*[\"'](https://[^\"']+\.supabase\.co)',
    "Supabase Key": r'(?:supabaseKey|anonKey|SUPABASE_ANON_KEY):\s*[\"'](eyJ[A-Za-z0-9_\\-]+\.[A-Za-z0-9_\\-]+\.[A-Za-z0-9_\\-]+)',
    "Heroku": r'[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}',
    "Generic Secret": r'(?:secret|password|token|key):\s*[\"']([^\"']{8,})',
}
```

## Batch Bundle Download + Grep

```python
import requests, re, json

base = "https://target.com"
html = requests.get(base).text

# Extract all JS URLs
js_urls = re.findall(r'src="([^"]*\.js)"', html)
for js_url in js_urls:
    if js_url.startswith("/"):
        js_url = base + js_url
    content = requests.get(js_url).text
    for name, pattern in patterns.items():
        matches = re.findall(pattern, content)
        for m in matches:
            if isinstance(m, tuple):
                m = m[0]
            if len(m) > 6:
                print(f"[{name}] {m[:80]}")
```

## Pitfalls

| Issue | Solution |
|-------|----------|
| Bundles too large | Use grep -Eo with specific patterns |
| Minified code (1 char names) | Use source maps for reconstruction |
| False positive matches | Validate keys by testing API endpoint |
| Rate limiting | Add delays between bundle downloads |

---

## Backend URL Discovery

JS bundles frequently leak production backend URLs, enabling direct API attacks bypassing CDN/WAF:

```bash
# Platform-specific backend URL patterns
grep -Eo 'https?://[a-zA-Z0-9.\-]+\.(fly\.dev|azurewebsites\.net|onrender\.com|vercel\.app|netlify\.app)[^"''' ]{0,40}' /tmp/*.js
grep -Eo 'https?://[a-zA-Z0-9.\-]+\.(supabase\.co|r2\.dev|blob\.vercel-storage\.com)[^"''' ]{0,40}' /tmp/*.js

# Edge function URLs
grep -Eo 'functions/v1/[a-zA-Z0-9_\-]+' /tmp/*.js

# Internal API paths
grep -Eo '["\x60]/api/v1/[a-zA-Z0-9_\-/]+["\x60]' /tmp/*.js
```

### Real Field Patterns
| Pattern | Platform | Example | Secret? |
|---------|----------|---------|---------|
| `*.fly.dev` | Fly.io | `ht-prod-backend.fly.dev` | ✅ Backend URL |
| `*.azurewebsites.net` | Azure | `consigpro-api-prod-...` | ✅ Backend URL |
| `*.onrender.com` | Render | `clickcity-api.onrender.com` | ✅ Backend URL |
| `*.supabase.co` | Supabase | `jxhvjufqtabpeieyhkgk.supabase.co` | ✅ Anon key is public; backend URL is intel |
| `*.r2.dev` | Cloudflare R2 | `pub-xxx.r2.dev` | ✅ Storage URL |
| `functions/v1/*` | Supabase Edge | `provision-openrouter-key` | ✅ Endpoint name |
| `dpl_*` | Vercel DPL | `dpl_BCoyPsxxYLZ...` | ❌ **NOT a secret** — public deploy ID |

## Verification

```bash
# Test Firebase API key
curl --max-time 30 --connect-timeout 10 -s "https://identitytoolkit.googleapis.com/v1/accounts:signUp?key=AIza..."
# Test Supabase anon key
curl --max-time 30 --connect-timeout 10 -s "https://PROJECT.supabase.co/rest/v1/users?limit=1" -H "apikey: ANON_KEY" -H "Authorization: Bearer ANON_KEY"
```

### Phase 5 — Source Map Exploitation

Recover full pre-compiled source code when `.js.map` files are left in production:

```bash
# Find .map files via Wayback Machine
curl --max-time 30 --connect-timeout 10 -s "https://web.archive.org/cdx/search/cdx?url=*.target.com/*&collapse=urlkey&output=text&fl=original&filter=original:.*\.js\.map$" \
  | sort -u > map_urls.txt

# Download and extract source
wget https://target.com/static/app.js.map
node -e "
const map = require('./app.js.map');
map.sources.forEach((src, i) => {
  const fs = require('fs');
  fs.writeFileSync(src.split('/').pop(), map.sourcesContent[i]);
});
print('Extracted ' + map.sources.length + ' source files');
"

# Quick check: does a JS file have an available map?
curl --max-time 30 --connect-timeout 10 -skI "https://target.com/static/app.js.map" | grep "200\|Content-Type"
```

### Phase 6 — Deep JS Crawling

Crawl JS files recursively for embedded URLs, APIs, and IPs:

```bash
# lazyegg — crawls JS files for links, APIs, IPs
python3 lazyegg.py https://target.com
python3 lazyegg.py https://target.com/js/auth.js

# Combine with waybackurls for deep coverage
waybackurls target.com \
  | grep '\.js$' \
  | awk -F '?' '{print $1}' \
  | sort -u \
  | xargs -I{} bash -c 'python3 lazyegg.py "{}" --js_urls --domains --ips' \
  > lazyegg_output.txt

# subjs — extract JS URLs from any URL list
cat all_urls.txt | subjs | tee js_files_full.txt
```

### Phase 7 — Per-File AI-Assisted Code Review

JS bundles are source code — even minified. A disciplined per-file (per-chunk) review finds what autonomous agents miss:

```bash
# 1. Download all JS chunks
curl --max-time 30 --connect-timeout 10 -sk "https://target.com" | grep -Eo 'src="[^"]+\.js[^"]*"' | \
  cut -d'"' -f2 | while read js; do
    curl --max-time 30 --connect-timeout 10 -sk "$js" -o "chunks/$(basename $js)"
  done

# 2. Per-chunk pattern review for dangerous sinks
for chunk in chunks/*.js; do
  echo "=== $chunk ==="
  # eval / new Function (arbitrary code execution)
  grep -Eon 'eval\s*\(|new\s+Function\s*\(' "$chunk"
  # Hardcoded API keys/secrets
  grep -Eon '(?:api[_-]?key|secret|token|password|bearer)\s*[:=]\s*["\x27][^"\x27]{8,}' "$chunk"
  # postMessage without origin check
  grep -Eon 'postMessage\s*\(' "$chunk"
  # Prototype pollution patterns
  grep -Eon '__proto__|constructor\.prototype' "$chunk"
  # Debug/test code in production
  grep -Eoin 'debug|test|staging|localhost' "$chunk"
  # Client-trusted flags
  grep -Eon '(?:isAdmin|isVip|isPremium|isModerator|role)\s*[=:]\s*true' "$chunk"
done > ai_review_findings.txt

# 3. Review findings — each is a CANDIDATE, not confirmed
grep -c "===" ai_review_findings.txt  # files reviewed
grep -c ":" ai_review_findings.txt     # candidate findings
```

Key insight: autonomous agents told "find bugs" in a whole codebase burn budget and miss things. A guaranteed per-file pass with fixed output structure produces repeatable hits. Each finding still needs manual PoC verification.