Skip to content

ShulkwiSEC/bb-huge / skills/curated/bug-bounty-workflow-funnel

bug-bounty-workflow-funnel

Verified

Implement the 5-stage Funnel workflow (Notes → Leads → Primitives → Findings → Reports) for structured bug bounty hunting. Based on Critical Thinking Bug Bounty Podcast Episode 166.

Bug Bounty Workflow Funnel

When to Use

  • When starting a new bug bounty program and need a structured data management system.
  • When drowning in raw notes and need to triage observations into actionable items.
  • When preparing to convert confirmed bugs into polished submission reports.
  • When working in a team and need a shared vocabulary for finding severity stages.

Prerequisites

  • A target program with defined scope (HackerOne / Bugcrowd / Intigriti / VDP)
  • A note-taking system (Obsidian, Notion, plain Markdown files, or similar)
  • Claude Code CLI for AI-assisted triage and report generation (optional but recommended)

Core Concept: The 5-Stage Funnel

"Structure your hacking data like a sales funnel. Raw observations at the top, polished reports at the bottom." — Critical Thinking Podcast, Ep. 166 [37:42]

flowchart TB
    N["📝 NOTES<br/>(All raw observations)"]
    L["🔍 LEADS<br/>(Interesting patterns)"]
    P["🔧 PRIMITIVES / GADGETS<br/>(Reusable techniques & entry points)"]
    F["🐛 FINDINGS<br/>(Confirmed vulnerabilities)"]
    R["📄 REPORTS<br/>(Submission-ready documentation)"]
    
    N --> L --> P --> F --> R
    
    style N fill:#1a1a2e,stroke:#e94560,color:#fff
    style L fill:#1a1a2e,stroke:#f9a826,color:#fff
    style P fill:#1a1a2e,stroke:#16c79a,color:#fff
    style F fill:#1a1a2e,stroke:#11999e,color:#fff
    style R fill:#1a1a2e,stroke:#6c5ce7,color:#fff

Workflow

Stage 1: Notes (Raw Observations)

Everything goes here. No filtering. No judgment. Write it down NOW, evaluate LATER.

Directory structure:

target-name/
├── notes/
│   ├── 2025-01-15-initial-recon.md
│   ├── 2025-01-15-auth-flow-analysis.md
│   ├── 2025-01-16-api-endpoint-mapping.md
│   └── 2025-01-16-js-source-review.md

Note template:

# [Date] [Topic]

## Context
- Target: [subdomain/endpoint/feature]
- Tool used: [Burp/Browser DevTools/ffuf/etc]
- Time spent: [duration]

## Observations
- [Observation 1]
- [Observation 2]

## Raw Data
[Paste interesting requests/responses/code snippets here]

## Gut Feeling
- [ ] Worth investigating further
- [ ] Dead end
- [ ] Need more context

What belongs in Notes:

  • HTTP request/response pairs that look unusual
  • JavaScript variables or API endpoints found in source code
  • Error messages, stack traces, verbose responses
  • Authentication token formats observed
  • Rate limit behaviour observations
  • Any deviation from expected application behaviour

Stage 2: Leads (Interesting Patterns)

Leads are notes that survived initial triage. Something here is worth deeper investigation.

Promotion criteria (Notes → Leads):

SignalExample
Reflected user inputParameter value appears in response body/headers
Inconsistent authorizationDifferent responses for same endpoint with different user roles
Verbose error messagesStack trace leaking framework version, file paths
Unprotected API endpoints/api/internal/ accessible without auth
Client-side security controlsJavaScript validation with no server-side enforcement
Unusual HTTP headersX-Debug: true, X-Powered-By: Express 4.17.1

Lead template:

# LEAD: [Short Description]

## Source Note
- Link: [notes/2025-01-15-auth-flow-analysis.md]

## Hypothesis
[What vulnerability might this lead to?]

## Evidence
[Specific request/response or code snippet proving this is worth pursuing]

## Next Steps
1. [Action item 1]
2. [Action item 2]

## Priority: [HIGH / MEDIUM / LOW]

Stage 3: Primitives / Gadgets (Reusable Techniques)

Primitives are common attack techniques or entry points you discover that can be reused across multiple targets. Think of them as building blocks for exploitation chains.

Examples:

PrimitiveDescriptionReuse Potential
Unsigned JWTApplication accepts alg: none JWTsAny endpoint using this auth
GraphQL Introspection__schema query returns full type systemEnumerate all mutations
Predictable ID sequenceUser IDs are sequential integersIDOR on any user-scoped endpoint
Open redirect on login/login?redirect= accepts any URLOAuth token theft chain
File upload → path traversalUpload filename is used in storage pathWrite to any directory

Primitive template:

# PRIMITIVE: [Technique Name]

## Description
[What is this technique and why does it work?]

## How to Detect
[Specific test to confirm this primitive exists on a target]

## Exploitation
[Step-by-step usage]

## Chain Potential
[What can this be combined with? SSRF + this = ? IDOR + this = ?]

## Seen On
- [Target 1] — [Date]
- [Target 2] — [Date]

Stage 4: Findings (Confirmed Vulnerabilities)

A finding is a CONFIRMED bug with a working proof-of-concept. No speculation — only reproducible exploitation.

Promotion criteria (Leads → Findings):

  • ✅ Reproducible exploit with specific steps
  • ✅ Clear security impact (data leak, account takeover, RCE, etc.)
  • ✅ Within program scope
  • ✅ Not a known/acceptable risk documented in program policy
  • ❌ NOT: "I think this might be exploitable" — that stays as a Lead

Finding template:

# FINDING: [Vulnerability Type] in [Component]

## Severity: [Critical / High / Medium / Low]
## CVSS 4.0: [Score] — [Vector String]

## Summary
[2-3 sentences: what is broken, how it is exploited, worst-case impact]

## Proof of Concept
[Exact curl command / Burp request / script that reproduces the bug]

## Impact
[Specific business impact with numbers if possible]

## Root Cause
[Why does this vulnerability exist? Missing validation? Broken access control?]

## Remediation
[Specific code fix, not generic advice]

## Status: [CONFIRMED / SUBMITTED / TRIAGED / RESOLVED / DUPLICATE]

Stage 5: Reports (Submission-Ready Documentation)

The final polished document submitted to the bug bounty platform. This is where money is made or lost.

Report quality directly correlates with bounty amount.

See the elite-report-writing skill for the full HackerOne-optimized report template.

Quick Report Checklist:

  • Title uses format: [VulnType] in [Component] Allows [BusinessImpact]
  • Summary is exactly 2-4 sentences
  • CVSS 4.0 vector is defensible
  • Steps to Reproduce are copy-paste deterministic
  • PoC is a runnable script/command with zero placeholders
  • Impact includes at least one real number
  • Remediation gives exact code fix (before/after)
  • CWE ID is correct

Claude Integration

Use Claude to automate transitions between stages:

Prompt to Claude:
"Review all files in notes/ and identify any that should be promoted to leads/.
For each promotion, create a lead file with hypothesis, evidence, and priority.
Explain why each note was or was not promoted."
Prompt to Claude:
"Take the finding in findings/idor-user-profiles.md and generate a 
HackerOne-ready report in reports/. Use the elite report template. 
Validate the CVSS score against the described impact."

Directory Structure for a Target

programs/
└── example-corp/
    ├── .claudemd           # Scope, policy, auth tokens (see ai-pair-hunting skill)
    ├── notes/              # Stage 1: Everything goes here
    ├── leads/              # Stage 2: Worth investigating
    ├── primitives/         # Stage 3: Reusable gadgets
    ├── findings/           # Stage 4: Confirmed bugs
    ├── reports/            # Stage 5: Submitted reports
    └── scripts/            # Custom automation for this target

Creativity Directive

IMPORTANT: The 5-stage funnel is a framework, not a cage. You are expected to adapt it:

  • Add sub-stages if a target is complex (e.g., Leads-High / Leads-Low).
  • Create cross-references between primitives and findings.
  • Auto-generate statistics (conversion rate: notes→findings).
  • Build a "dormant leads" watchlist for findings that need updated tool versions.

Think like an attacker. Adapt. Improvise.

🔵 Blue Team

  • Deploy robust WAF rules to detect anomalies.
  • Monitor logs for unusual access patterns.

📚 Shared Resources

For cross-cutting methodology applicable to all vulnerability classes, see:

References

planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies.Stars: 108kUpdated Claude CodeCodexVerifiedInstall
meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要".Stars: 108kUpdated Claude CodeCodexVerifiedInstall