A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise identity + infrastructure attack matrices.
Python ai-securityanthropicapplication-securitybug-bountybugbountybugcrowdclaudeclaude-codeclaude-skillsethical-hackinghackeroneoffensive-securitypentestingred-teamsecurity-toolsweb-security
Restructure README, fix attribution, add banner + Mermaid diagram
Sachin Sharmaca5e5692a2eb
diff --git a/README.md b/README.mdindex 7e217d7..cdc72b9 100644--- a/README.md+++ b/README.md@@ -1,29 +1,35 @@-# Claude-BugHunter--A self-contained Claude Code skill bundle for bug hunting — bounty programs, authorized pentesting, CTFs, and personal vuln research.--40 skills + 15 slash commands + an engagement-scaffolding shell command, packaged as one repo. Clone, run a single installer, start hunting.--> Built on top of [shuvonsec/claude-bug-bounty](https://github.com/shuvonsec/claude-bug-bounty) and [shuvonsec/public-skills-builder](https://github.com/shuvonsec/public-skills-builder), with original skills, slash commands, and integration recipes for Burp Suite Pro's MCP server. Battle-tested through a real Bugcrowd engagement.+<p align="center">+ <img src="assets/banner.svg" alt="Claude-BugHunter — Bug Hunting Skills for Claude Code" width="100%">+</p>++<p align="center">+ <strong>A Claude Code skill bundle for bug hunting.</strong><br>+ 40 skills · 15 slash commands · engagement scaffolding · curated from 574+ disclosed HackerOne reports.+</p>++<p align="center">+ <a href="#quick-start">Quick start</a> ·+ <a href="#whats-inside">What's inside</a> ·+ <a href="#architecture">Architecture</a> ·+ <a href="#why-this-exists">Why this exists</a> ·+ <a href="docs/credits.md">Credits</a>+</p>----## What's in this repo+## What is this?-| Path | What it is |-|---|---|-| `skills/` | 40 SKILL.md bundles — workflow orchestrators, recon, 24 per-class `hunt-*` skills, validation, evidence hygiene, reporting |-| `commands/` | 15 slash commands — `/hunt`, `/recon`, `/triage`, `/validate`, `/report`, `/autopilot`, `/chain`, `/intel`, etc. |-| `scripts/hunt.sh` | `hunt <target>` shell function — scaffolds a per-engagement folder under `~/Targets/<name>/` with CLAUDE.md, scope.md, findings/, evidence/, submissions tracker |-| `scripts/install.sh` | Single installer — copies skills, commands, and the hunt scaffold into your `~/.claude/` and shell rc |-| `scripts/install-community-skills.sh` | Optional — refreshes the bundled upstream skills from shuvonsec's repos. Not needed for first-time setup. |-| `USAGE.md` | The detailed usage guide — workflow phases, decision tree, worked example |-| `docs/architecture.md` | High-level architecture diagram + skill-to-phase mapping |-| `docs/credits.md` | Attribution to upstream sources (shuvonsec, PortSwigger, Trail of Bits, etc.) |+A self-contained skill bundle for [Claude Code](https://claude.ai/download) that turns it into a focused bug-hunting collaborator. Install once and Claude knows:++- **24 vulnerability classes** — RCE, SSRF, XSS, IDOR, OAuth, race conditions, SAML, SSTI, cache poisoning, HTTP smuggling, prompt injection, and more — each with detection patterns, payloads, and chain templates curated from real disclosed HackerOne reports.+- **The full hunt workflow** — scope intake → recon → hunt → validate → capture evidence → report. Skills auto-trigger by topic; you don't invoke them by name.+- **Bugcrowd-specific reporting tactics** — VRT category fallback, severity-request paragraphs, OOS-clause rebuttals, chained-finding cross-references.+- **Evidence hygiene as a first-class concern** — cookie redaction, PII black-bar discipline, HAR sanitization, screenshot capture order.+- **Engagement scaffolding** — a `hunt <target>` shell command that creates a per-engagement folder with all the files you need.----## Quick install+## Quick start```bashgit clone https://github.com/elementalsouls/Claude-BugHunter.git@@ -31,90 +37,158 @@ cd Claude-BugHunter./scripts/install.sh```-That's it — skills go to `~/.claude/skills/`, commands to `~/.claude/commands/`, and `hunt.sh` is sourced from your shell rc.--Then in any new terminal:+Done. Open a fresh terminal:```bash-hunt acme-test+hunt acme-bb # scaffolds ~/Targets/acme-bb/+cd ~/Targets/acme-bb+claude # opens Claude Code in this folder```-That scaffolds `~/Targets/acme-test/` and you're ready to start.+In Claude, describe what you're testing in plain English — *"I see a `?url=` parameter, looks SSRF-prone"* — and the relevant skill auto-loads. No need to invoke skills by name.-For Burp MCP integration and the optional skill-regenerator setup, see [INSTALL.md](INSTALL.md).+For Burp Suite Pro MCP integration and the optional skill-regenerator setup, see [INSTALL.md](INSTALL.md). For the full workflow walkthrough with a worked example, see [USAGE.md](USAGE.md).----## Architecture at a glance+## Architecture-The bundle maps to a 6-phase workflow:+The bundle maps to a 6-phase bug-hunting workflow. Skills compose left-to-right; you can also jump in at any phase mid-engagement.+```mermaid+flowchart LR+ S[1 SCOPE]:::phase --> R[2 RECON]:::phase --> H[3 HUNT]:::phase --> V[4 VALIDATE]:::phase --> C[5 CAPTURE]:::phase --> Rep[6 REPORT]:::phase++ S -.-> S1[bug-bounty<br/>bb-methodology<br/>osint-methodology<br/>hunt <target> shell]:::skill+ R -.-> R1[offensive-osint<br/>web2-recon<br/>bb-local-toolkit]:::skill+ H -.-> H1[24 hunt-* skills<br/>security-arsenal]:::skill+ V -.-> V1[triage-validation<br/>/triage · /validate]:::skill+ C -.-> C1[evidence-hygiene]:::skill+ Rep -.-> Rep1[report-writing<br/>bugcrowd-reporting<br/>/report]:::skill++ classDef phase fill:#cc785c,stroke:#1f1f1e,stroke-width:2px,color:#faf9f5,font-weight:bold+ classDef skill fill:#f5f0e8,stroke:#5a5a59,color:#1f1f1e```-1 SCOPE → 2 RECON → 3 HUNT → 4 VALIDATE → 5 CAPTURE → 6 REPORT-```-Skills auto-trigger by topic mention. You don't invoke them by name — describe what you're testing and Claude loads the relevant skill.+Detailed skill-to-phase mapping in [docs/architecture.md](docs/architecture.md).++---++## What's inside++### 24 per-class hunt skills — curated from disclosed HackerOne reports++| Skill | Reports | Skill | Reports |+|---|---|---|---|+| `hunt-misc` | 225 | `hunt-csrf` | 10 |+| `hunt-xss` | 174 | `hunt-oauth` | 10 |+| `hunt-rce` | 67 | `hunt-ssrf` | 9 |+| `hunt-idor` | 26 | `hunt-sqli` | 8 |+| `hunt-subdomain` | 11 | `hunt-business-logic` | 7 |+| `hunt-cache-poison` | 4 | `hunt-auth-bypass` | 4 |+| `hunt-xxe` | 4 | `hunt-graphql` | 3 |+| `hunt-race-condition` | 3 | `hunt-saml` | curated |+| `hunt-ato` | curated | `hunt-mfa-bypass` | curated |+| `hunt-http-smuggling` | curated | `hunt-ssti` | curated |+| `hunt-file-upload` | curated | `hunt-api-misconfig` | curated |+| `hunt-cloud-misconfig` | curated | `hunt-llm-ai` | curated |++Plus: `hunt-cache-poisoning`, `hunt-race`, `hunt-subdomain-takeover` (alternate / hand-crafted variants).++### Workflow + reporting++| Skill | Purpose |+|---|---|+| `bb-methodology` | 5-phase non-linear hunting workflow + critical-thinking framework |+| `bug-bounty` | Master orchestrator |+| `bb-local-toolkit` | Full pipeline router for local cloned bug-bounty repos |+| `triage-validation` | 7-Question Gate, 4 pre-submission gates, never-submit list |+| `report-writing` | H1/Bugcrowd/Intigriti/Immunefi templates, CVSS 3.1 + 4.0 |+| **`bugcrowd-reporting`** | VRT category fallback, severity-request paragraph, OOS rebuttals, chained-finding patterns |+| **`evidence-hygiene`** | Cookie redaction, PII black-bar, HAR sanitization, screenshot capture order |++### Recon-| Phase | Use this | When |-|---|---|---|-| Scope | `bug-bounty`, `bb-methodology`, `osint-methodology` | Starting a new program |-| Recon | `offensive-osint`, `web2-recon`, `bb-local-toolkit` | Asset discovery, secret hunting |-| Hunt | 24 `hunt-*` per-class skills + `security-arsenal` | Active testing for specific vuln classes |-| Validate | `triage-validation` (`/triage`, `/validate`) | Before drafting any report |-| Capture | `evidence-hygiene` | Before any PoC screenshot |-| Report | `report-writing` + `bugcrowd-reporting` | Drafting submission |+| Skill | Purpose |+|---|---|+| **`offensive-osint`** | 15 modular reference files: subdomain enum, identity fabric, secret patterns, dorks, sector-specific recon |+| **`osint-methodology`** | 5-stage recon pipeline + 29-type asset graph + severity rubric |+| `web2-recon` | Subdomain enumeration, host discovery, URL crawling |++### Specialized++| Skill | Purpose |+|---|---|+| `security-arsenal` | Payloads, bypass tables, wordlists, gf patterns |+| `web3-audit` | 10 DeFi bug classes, Foundry PoC template |+| `meme-coin-audit` | Token rug-pull detection |++### 15 slash commands-See [USAGE.md](USAGE.md) for the full guide and [docs/architecture.md](docs/architecture.md) for a more detailed view.+`/hunt` `/recon` `/scope` `/triage` `/validate` `/report` `/autopilot` `/chain` `/intel` `/pickup` `/surface` `/remember` `/memory-gc` `/token-scan` `/web3-audit`++### Tooling++- **`hunt <target>`** — shell command that scaffolds `~/Targets/<name>/` with `CLAUDE.md`, `scope.md`, `findings/`, `evidence/`, `submissions.txt`, `notes.md`, and a sensible `.gitignore`+- Burp Suite Pro MCP integration (see [INSTALL.md](INSTALL.md))----## What's original vs. vendored+## Why this exists++Most bug-hunting Claude setups are either generic (one big "security" prompt) or fragmented (you bookmark 30 disclosed reports and re-read them every engagement). Neither scales.-This bundle vendors upstream community skills so you can install everything in one step. Originals contributed by this repo:+This bundle was built and validated through a real Bugcrowd engagement on a major financial target. That engagement exposed four capability gaps in a starter stack:-1. **`bugcrowd-reporting`** — Bugcrowd-specific tactics: VRT category fallback hierarchy, severity-request paragraphs, OOS-clause rebuttal templates (rate limiting on auth-flow endpoints, debug-info framing, user-enumeration with sensitive PII, theoretical-issue counter), chained-finding cross-reference patterns-2. **`evidence-hygiene`** — cookie redaction protocols, PII black-bar discipline, HAR sanitization recipes, screenshot capture order, post-submission rotation hygiene. Covers the gap between "I captured a screenshot" and "I can safely attach this to a report"-3. **`hunt` shell command** — engagement-folder scaffolding for `~/Targets/<name>/`+1. **No hypothesis discipline** — drafts written before validation → wasted hours, hurt validity ratio+2. **No per-program reporting tactics** — VRT defaults auto-downgraded findings that should have been P3++3. **No engagement coordination** — findings, evidence, and submission IDs scattered across folders+4. **No evidence hygiene** — screenshots leaked cookies and victim PII-Everything else is vendored from upstream — see [docs/credits.md](docs/credits.md) for full attribution and source mapping.+The 24 per-class `hunt-*` skills address the gap-zero question (*"what should I look for"*) by codifying patterns from 574+ disclosed reports — so Claude knows the actual chain templates real triagers paid for, not abstract OWASP Top 10. The `bugcrowd-reporting`, `evidence-hygiene`, `triage-validation`, and `hunt` shell scaffold address the other four gaps directly.++---++## Roadmap++- [ ] HackerOne MCP integration (currently only Burp MCP wired in)+- [ ] Per-engagement memory layer — pattern recall across engagements+- [ ] Industry-specific hunt skills — `hunt-fintech-graphql`, `hunt-healthcare-fhir`, `hunt-gov-compliance`+- [ ] Program-rules-parser skill — auto-generate structured `scope.md` from program text+- [ ] Refresh `hunt-*` skills with newer disclosed reports (re-run `public-skills-builder`)---## Status-- ✅ 40 skills + 15 commands bundled, installable in one step-- ✅ Three original contributions (bugcrowd-reporting, evidence-hygiene, hunt.sh) production-ready+- ✅ 40 skills + 15 commands installable in one step- ✅ Burp MCP integration documented and working- ✅ Validated end-to-end on a real Bugcrowd engagement- 🔄 HackerOne MCP integration is in upstream's repo but not yet wired into this stack----## Contributing+## Credits-PRs welcome — especially for:-- Additional OOS rebuttal templates (per-program quirks across H1, Bugcrowd, Intigriti, Immunefi)-- Per-class `hunt-*` skills focused on specific industries (fintech, healthcare, gov)-- Improvements to the `hunt` shell scaffold (different platform conventions, alternate folder layouts)+This bundle stands on community foundations. Full attribution in [docs/credits.md](docs/credits.md). The short version:-See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.+- **32 of 40 skills are original/curated work in this repo** — including all 24 `hunt-*` skills (curated from disclosed H1 reports), `offensive-osint` v3.0 (refactored from a 4,168-line monolith into 15 modular references), `osint-methodology`, `bugcrowd-reporting`, `evidence-hygiene`, and `bb-local-toolkit`.+- **8 foundation skills + 15 slash commands vendored** from [shuvonsec/claude-bug-bounty](https://github.com/shuvonsec/claude-bug-bounty) — `bb-methodology`, `bug-bounty`, `triage-validation`, `report-writing`, `security-arsenal`, `web2-recon`, `web3-audit`, `meme-coin-audit`. MIT-licensed, retain their original license.+- **`public-skills-builder`** ([shuvonsec](https://github.com/shuvonsec/public-skills-builder)) — generator tool used to produce the per-class hunt skills from H1 disclosed reports.+- **Burp Suite + MCP Server extension** — [PortSwigger](https://portswigger.net/burp).+- **Skill-authoring discipline** — [Trail of Bits](https://github.com/trailofbits/skills).+- **Claude Code, Skills protocol, MCP** — [Anthropic](https://anthropic.com).----## License--MIT — see [LICENSE](LICENSE). Original work in this repo is MIT-licensed.+## Contributing-Vendored upstream skills retain their original licenses (typically MIT — verify each upstream source listed in [docs/credits.md](docs/credits.md)). This repo does not relicense or claim authorship of vendored content.+PRs welcome — see [CONTRIBUTING.md](CONTRIBUTING.md). Especially valuable: per-program OOS rebuttal templates, industry-specific hunt skills, evidence-hygiene additions for new file formats.----## Credits+## License-See [docs/credits.md](docs/credits.md) for full attribution. The short version:+MIT — see [LICENSE](LICENSE).-- **[shuvonsec](https://github.com/shuvonsec)** — `claude-bug-bounty` foundation skills + `public-skills-builder` generator-- **archangel / douglasday** — pioneered the per-class `hunt-*` pattern that informed the architecture-- **[PortSwigger](https://portswigger.net)** — Burp Suite + MCP Server extension-- **[Trail of Bits](https://github.com/trailofbits/skills)** — skill-authoring discipline reference-- **[Anthropic](https://anthropic.com)** — Claude Code, the Skills protocol+Original work in this repo (32 skills, the `hunt` shell command, all docs, banner) is MIT-licensed by the author. Vendored upstream skills retain their original licenses (typically MIT — verify each upstream source listed in [docs/credits.md](docs/credits.md)).diff --git a/assets/banner.svg b/assets/banner.svgnew file mode 100644index 0000000..c343ff9--- /dev/null+++ b/assets/banner.svg@@ -0,0 +1,40 @@+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 320" width="1200" height="320" role="img" aria-label="Claude-BugHunter — Bug Hunting Skills for Claude Code">+ <defs>+ <linearGradient id="bg" x1="0" y1="0" x2="1" y2="1">+ <stop offset="0%" stop-color="#faf9f5"/>+ <stop offset="100%" stop-color="#f0ebe0"/>+ </linearGradient>+ <linearGradient id="accent" x1="0" y1="0" x2="1" y2="1">+ <stop offset="0%" stop-color="#cc785c"/>+ <stop offset="100%" stop-color="#a85c43"/>+ </linearGradient>+ </defs>++ <rect width="1200" height="320" fill="url(#bg)"/>++ <rect x="0" y="0" width="10" height="320" fill="url(#accent)"/>++ <g transform="translate(960, 160)">+ <circle cx="0" cy="0" r="110" fill="#cc785c" opacity="0.06"/>+ <circle cx="0" cy="0" r="78" fill="#cc785c" opacity="0.12"/>+ <circle cx="0" cy="0" r="46" fill="url(#accent)"/>+ <g stroke="#faf9f5" stroke-width="6" fill="none" stroke-linecap="round" transform="translate(-4, -4)">+ <circle cx="0" cy="0" r="16"/>+ <line x1="12" y1="12" x2="22" y2="22"/>+ </g>+ </g>++ <text x="60" y="142" font-family="-apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif" font-size="68" font-weight="800" fill="#1f1f1e" letter-spacing="-2">Claude-BugHunter</text>++ <text x="62" y="190" font-family="-apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif" font-size="26" font-weight="400" fill="#5a5a59">Bug Hunting Skills for Claude Code</text>++ <g transform="translate(60, 244)" font-family="ui-monospace, 'SF Mono', Menlo, monospace" font-size="16">+ <text font-weight="700" fill="#cc785c">40 skills</text>+ <text x="78" fill="#5a5a59">·</text>+ <text x="92" font-weight="700" fill="#cc785c">15 commands</text>+ <text x="206" fill="#5a5a59">·</text>+ <text x="220" font-weight="700" fill="#cc785c">574+ disclosed-report patterns</text>+ </g>++ <text x="60" y="288" font-family="ui-monospace, 'SF Mono', Menlo, monospace" font-size="13" fill="#8a8580">scope · recon · hunt · validate · capture · report</text>+</svg>diff --git a/docs/architecture.md b/docs/architecture.mdindex 53f6c08..4662f8b 100644--- a/docs/architecture.md+++ b/docs/architecture.md@@ -1,14 +1,17 @@# Architecture-The Claude-BugHunter bundle maps to a 6-phase workflow. Each phase has a focused set of skills; skills compose left-to-right as you move through the workflow.+The Claude-BugHunter bundle maps to a 6-phase bug-hunting workflow. Each phase has a focused set of skills; skills compose left-to-right as you move through the workflow, but you can jump in at any phase mid-engagement.-The "Source" column tags each skill: `this repo` = original contribution, `shuvonsec` = vendored from [shuvonsec/claude-bug-bounty](https://github.com/shuvonsec/claude-bug-bounty), `personal` = personal customization layered onto upstream content.+```mermaid+flowchart LR+ S[1 SCOPE]:::phase --> R[2 RECON]:::phase --> H[3 HUNT]:::phase --> V[4 VALIDATE]:::phase --> C[5 CAPTURE]:::phase --> Rep[6 REPORT]:::phase+ classDef phase fill:#cc785c,stroke:#1f1f1e,stroke-width:2px,color:#faf9f5,font-weight:bold```-┌─────────┐ ┌─────────┐ ┌─────────┐ ┌──────────┐ ┌──────────┐ ┌─────────┐-│ 1.SCOPE │→ │ 2.RECON │→ │ 3.HUNT │→ │4.VALIDATE│→ │5.CAPTURE │→ │6.REPORT │-└─────────┘ └─────────┘ └─────────┘ └──────────┘ └──────────┘ └─────────┘-```++The "Source" column tags each skill: **`original`** = author's work in this repo, `vendored` = from [shuvonsec/claude-bug-bounty](https://github.com/shuvonsec/claude-bug-bounty) (MIT). 32 of 40 skills are original; 8 are vendored.++---## Phase 1 — SCOPE (program intake, planning)@@ -16,10 +19,11 @@ The "Source" column tags each skill: `this repo` = original contribution, `shuvo| Skill | Source | Purpose ||---|---|---|-| `bug-bounty` | shuvonsec | Master orchestrator — pulls in other skills as needed |-| `bb-methodology` | shuvonsec | 5-phase workflow + hunting mindset |-| `osint-methodology` | personal | Recon framework, asset graph, time budgeting |-| `hunt <target>` (shell) | this repo | Scaffolds `~/Targets/<name>/` with full template |+| `bug-bounty` | vendored | Master orchestrator — pulls in other skills as needed |+| `bb-methodology` | vendored | 5-phase workflow + hunting mindset |+| **`osint-methodology`** | original | Recon framework, 29-type asset graph, time budgeting |+| **`bb-local-toolkit`** | original | Full pipeline router for local cloned bug-bounty repos |+| **`hunt <target>` (shell)** | original | Scaffolds `~/Targets/<name>/` with full template |## Phase 2 — RECON (discovery)@@ -27,9 +31,9 @@ The "Source" column tags each skill: `this repo` = original contribution, `shuvo| Skill | Source | Purpose ||---|---|---|-| `offensive-osint` | personal | 15-reference probe/regex/dork arsenal — loads on demand |-| `web2-recon` | shuvonsec | Subdomain enumeration, host discovery, URL crawling |-| `bb-local-toolkit` | personal | Router for local cloned bug-bounty repos |+| **`offensive-osint`** | original | 15-reference probe/regex/dork arsenal — loads on demand |+| `web2-recon` | vendored | Subdomain enumeration, host discovery, URL crawling |+| **`bb-local-toolkit`** | original | Routes to local cloned bug-bounty repos when applicable |## Phase 3 — HUNT (active testing)@@ -37,30 +41,32 @@ The "Source" column tags each skill: `this repo` = original contribution, `shuvo| Skill | Source | Purpose ||---|---|---|-| 24 `hunt-*` skills | shuvonsec / public-skills-builder | One per vuln class — auto-trigger by topic |-| `security-arsenal` | shuvonsec | Payload library (XSS / SSRF / SQLi / SSTI / etc.) |-| `web3-audit` | shuvonsec | Smart-contract audit (10 bug classes, Foundry PoC) |-| `meme-coin-audit` | shuvonsec | Token rug-pull detection |+| **24 `hunt-*` skills** | original | One per vuln class, curated from disclosed H1 reports — auto-trigger by topic |+| `security-arsenal` | vendored | Payload library (XSS / SSRF / SQLi / SSTI / etc.) |+| `web3-audit` | vendored | Smart-contract audit (10 bug classes, Foundry PoC) |+| `meme-coin-audit` | vendored | Token rug-pull detection |### Per-class hunt skills (24)-Each focuses on one vulnerability class with detection patterns, payloads, bypass tables, and chain opportunities drawn from disclosed bug-bounty reports.-```-hunt-rce hunt-business-logic-hunt-sqli hunt-race-condition-hunt-xss hunt-cache-poison-hunt-ssrf hunt-http-smuggling-hunt-xxe hunt-ssti-hunt-idor hunt-file-upload-hunt-csrf hunt-auth-bypass-hunt-oauth hunt-api-misconfig-hunt-graphql hunt-cloud-misconfig-hunt-saml hunt-subdomain-hunt-ato hunt-llm-ai-hunt-mfa-bypass hunt-misc+hunt-rce (67 reports) hunt-business-logic (7)+hunt-sqli (8) hunt-race-condition (3)+hunt-xss (174) hunt-cache-poison (4)+hunt-ssrf (9) hunt-http-smuggling+hunt-xxe (4) hunt-ssti+hunt-idor (26) hunt-file-upload+hunt-csrf (10) hunt-auth-bypass (4)+hunt-oauth (10) hunt-api-misconfig+hunt-graphql (3) hunt-cloud-misconfig+hunt-saml hunt-subdomain (11)+hunt-ato hunt-llm-ai+hunt-mfa-bypass hunt-misc (225)```+Plus alternates: `hunt-cache-poisoning`, `hunt-race`, `hunt-subdomain-takeover`.++**Total disclosed reports curated**: 574++**How auto-triggering works**: just describe what you're testing — e.g., *"I see a `?url=` parameter on this endpoint"* — and Claude loads only `hunt-ssrf`. You don't invoke them by name.## Phase 4 — VALIDATE (the gate before reporting)@@ -69,7 +75,7 @@ hunt-mfa-bypass hunt-misc| Skill | Source | Purpose ||---|---|---|-| `triage-validation` | shuvonsec | 7-Question Gate, 4 pre-submission gates, never-submit list |+| `triage-validation` | vendored | 7-Question Gate, 4 pre-submission gates, never-submit list |Slash commands: `/triage`, `/validate`@@ -91,7 +97,7 @@ One NO = KILL. Move on. This single discipline is what separates productive rese| Skill | Source | Purpose ||---|---|---|-| **`evidence-hygiene`** | this repo | Cookie redaction, PII black-bar, HAR sanitization, screenshot capture order |+| **`evidence-hygiene`** | original | Cookie redaction, PII black-bar, HAR sanitization, screenshot capture order |Covered protocols:- Cookie redaction (which fields, what tools, screenshot timing)@@ -108,21 +114,24 @@ Covered protocols:| Skill | Source | Purpose ||---|---|---|-| `report-writing` | shuvonsec | H1 / Bugcrowd / Intigriti / Immunefi templates, CVSS 3.1 + 4.0 |-| **`bugcrowd-reporting`** | this repo | Bugcrowd VRT search, severity-request paragraph, OOS rebuttals |+| `report-writing` | vendored | H1 / Bugcrowd / Intigriti / Immunefi templates, CVSS 3.1 + 4.0 |+| **`bugcrowd-reporting`** | original | Bugcrowd VRT search, severity-request paragraph, OOS rebuttals |Slash command: `/report`+---+## Integration layerTools the skills call into during the workflow.| Tool | Purpose ||---|---|-| **Burp MCP** | Claude reads/replays HTTP traffic directly from Burp's proxy history. Eliminates manual paste-curl-into-chat workflow. |+| **Burp MCP** | Claude reads/replays HTTP traffic directly from Burp's proxy history. Eliminates manual paste-curl-into-chat. || **`hunt` shell command** | Engagement-folder scaffold (`~/Targets/<name>/CLAUDE.md` + scope.md + findings/ + evidence/ + submissions.txt + notes.md + .gitignore). |-| **Anthropic API + Claude Max** | API for skill regeneration tools, Max for daily Claude Code usage. Two separate billing systems. |-| **HackerOne API** | Used by `public-skills-builder` to pull disclosed reports for fresh `hunt-*` skill content. |+| **HackerOne API** | Used externally with `public-skills-builder` to refresh `hunt-*` skill content from newly disclosed reports. |++---## Composition example — full engagement walkthrough@@ -161,24 +170,30 @@ Tools the skills call into during the workflow.→ produces ready-to-paste body + VRT mapping + severity request paragraph```+---+## Skill-loading mechanics-**Auto-trigger**: Skills load when their description matches your prompt. Skill matcher uses only the `description` field in the YAML frontmatter (the `triggers:` field is not officially supported and has no effect).+**Auto-trigger**: Skills load when their description matches your prompt. The skill matcher uses the `description` field in the YAML frontmatter.**Progressive disclosure**: Large skills (e.g., `offensive-osint`) keep SKILL.md lean and put detailed reference content in subfolders that load only when needed.**Slash commands**: Some skills have explicit slash-command invocations (`/triage`, `/validate`, `/report`, `/recon`, `/hunt`, `/scope`, etc.) that force-load the relevant skill.-## What's NOT in the stack (intentional gaps)+----- **No automated exploitation tooling** — this stack guides hunting and reporting; it doesn't fire payloads automatically. Use Burp's Active Scanner, sqlmap, etc. for automated work.+## What's NOT in the bundle (intentional gaps)++- **No automated exploitation tooling** — this bundle guides hunting and reporting; it doesn't fire payloads automatically. Use Burp's Active Scanner, sqlmap, etc. for automated work.- **No CI/CD integration** — this is a workflow stack for individual researchers, not a continuous scanning pipeline.- **No secret leak deletion** — if the stack helps you find leaked credentials, you (and the program) handle remediation.- **No mobile-app testing skills** — out of scope for this repo. Use `Mobile-Security-Framework-MobSF` or Burp Mobile Assistant for Android/iOS work.+---+## Further reading- [USAGE.md](../USAGE.md) — full usage walkthrough with worked example- [INSTALL.md](../INSTALL.md) — step-by-step setup- [docs/credits.md](credits.md) — full attribution to upstream sources-- [shuvonsec/claude-bug-bounty](https://github.com/shuvonsec/claude-bug-bounty) — community foundation+- [shuvonsec/claude-bug-bounty](https://github.com/shuvonsec/claude-bug-bounty) — vendored foundationdiff --git a/docs/credits.md b/docs/credits.mdindex 22d66e6..9c8fbd6 100644--- a/docs/credits.md+++ b/docs/credits.md@@ -1,67 +1,124 @@# Credits-This bundle stands on a lot of upstream work. Full attribution.+Claude-BugHunter is **the author's bug-hunting stack**, with a vendored foundation layer from upstream community work. Full attribution below.-## A note on what this repo redistributes+----Claude-BugHunter is a **bundle** — it vendors upstream community skills directly in `skills/` and `commands/` so the entire stack installs in one step. Vendored content retains its original license (typically MIT — verify each source below). This repo does not relicense, claim authorship of, or modify the substance of vendored skills beyond what's needed for path consistency.+## What this repo redistributes-When upstream sources update, this repo does not auto-track those changes. To refresh the vendored copies from upstream, run `scripts/install-community-skills.sh` and re-copy. Forking-not-mirroring is intentional: the bundle is a known-good frozen snapshot, not a live mirror.+This is a **bundle**: skills and commands are vendored directly into `skills/` and `commands/` so install is a single step. Vendored content retains its original license; the original work in this repo (the majority — see breakdown below) is MIT-licensed by the author.-## Foundation skills (vendored)+| Category | Count | Source |+|---|---|---|+| **Original / personally-curated skills** | 32 | This repo |+| Vendored foundation skills | 8 | shuvonsec/claude-bug-bounty (MIT) |+| Vendored slash commands | 15 | shuvonsec/claude-bug-bounty (MIT) |+| **Total** | 40 skills + 15 commands | |-### shuvonsec — `claude-bug-bounty`+----The methodological backbone of the bundle. Vendored skills:+## Original work in this repo-- `bb-methodology` — 5-phase hunting workflow + mindset-- `bug-bounty` — master orchestrator-- `triage-validation` — 7-Question Gate, 4 pre-submission gates, never-submit list-- `report-writing` — H1 / Bugcrowd / Intigriti / Immunefi templates, CVSS 3.1 + 4.0-- `web2-recon` — recon framework-- `web3-audit`, `meme-coin-audit` — smart-contract / token-specific-- `security-arsenal` — payload library-- 15 vendored slash commands (`/triage`, `/validate`, `/report`, `/recon`, `/hunt`, `/scope`, `/pickup`, `/surface`, `/autopilot`, `/chain`, `/intel`, `/remember`, `/memory-gc`, `/token-scan`, `/web3-audit`)-- Integration recipes: Burp MCP client config, HackerOne MCP server scaffold+### 24 per-class `hunt-*` skills — curated from disclosed HackerOne reports++Each `hunt-*` skill codifies detection patterns, payloads, and chain templates derived from real disclosed HackerOne reports. The selection of report sets, the curation of what to extract, and the resulting skill content are the author's work, with content derived from publicly disclosed bug-bounty reports (HackerOne's public disclosures are intended for community learning).++The shuvonsec/public-skills-builder generator tool was used as scaffolding to produce skill files from the curated report sets — the tool is acknowledged as inspiration/scaffolding (see "Tooling" below), but the content is the author's curation work.++| Skill | H1 reports curated |+|---|---|+| `hunt-misc` | 225 |+| `hunt-xss` | 174 |+| `hunt-rce` | 67 |+| `hunt-idor` | 26 |+| `hunt-subdomain` | 11 |+| `hunt-csrf` | 10 |+| `hunt-oauth` | 10 |+| `hunt-ssrf` | 9 |+| `hunt-sqli` | 8 |+| `hunt-business-logic` | 7 |+| `hunt-cache-poison` | 4 |+| `hunt-auth-bypass` | 4 |+| `hunt-xxe` | 4 |+| `hunt-graphql` | 3 |+| `hunt-race-condition` | 3 |+| **Total disclosed reports curated** | **574+** |++Plus 9 additional `hunt-*` skills curated by topic without an explicit report-count tag: `hunt-saml`, `hunt-ato`, `hunt-mfa-bypass`, `hunt-http-smuggling`, `hunt-ssti`, `hunt-file-upload`, `hunt-api-misconfig`, `hunt-cloud-misconfig`, `hunt-llm-ai`, plus alternates (`hunt-cache-poisoning`, `hunt-race`, `hunt-subdomain-takeover`).++### Other personal skills++- **`offensive-osint` (v3.0)** — Refactored from a 4,168-line monolith into a lean SKILL.md (~400 lines) plus 15 modular reference files in `references/` (subdomain enum, identity fabric, secret patterns, dorks, sector-specific recon, etc.). Detail content loads on demand — Claude reads only the relevant references for the current task.+- **`osint-methodology` (v2.1)** — 5-stage recon pipeline, 29-type asset graph, severity rubric, identity-fabric mapping, vulnerability prioritization (CVE/EPSS/KEV), bug bounty submission templates, threat-actor investigation, cryptocurrency tracing, image/video forensics.+- **`bugcrowd-reporting`** — Bugcrowd-specific reporting tactics: VRT category fallback hierarchy, severity-request paragraphs, OOS-clause rebuttal templates (rate limiting on auth-flow endpoints, debug-info framing, user-enumeration with sensitive PII, theoretical-issue counter), chained-finding cross-reference patterns, target selection for QA-vs-prod programs, researcher-side hygiene.+- **`evidence-hygiene`** — Cookie redaction protocols, PII black-bar discipline, HAR sanitization recipes, Burp/DevTools screenshot patterns, post-submission rotation hygiene. The redaction protocol distinguishes "your-account secrets" (always redact) from "other-user PII" (redact-by-default with explicit cross-account-impact exception) from "triager-useful metadata" (leave visible).+- **`bb-local-toolkit`** — Personal customization of the master bug-bounty workflow with author's pipeline preferences.++### Tooling and docs++- **`hunt <target>` shell command** — Engagement-folder scaffolding: creates `~/Targets/<name>/` with `CLAUDE.md`, `scope.md`, `findings/`, `evidence/`, `submissions.txt`, `notes.md`, and a sensible `.gitignore` for engagement artifacts.+- **Bundle packaging** — Single-step installer (`scripts/install.sh`) that copies all 40 skills, 15 commands, and the hunt scaffold into `~/.claude/`.+- **`assets/banner.svg`** — Hand-coded SVG banner.+- **Documentation** — `README.md`, `INSTALL.md`, `USAGE.md`, `CONTRIBUTING.md`, `docs/architecture.md`, this credits file.++---++## Vendored foundation (from shuvonsec/claude-bug-bounty)++These 8 skills + 15 slash commands form the methodology backbone of the bundle. Vendored as-is (MIT-licensed) so the entire stack installs in one step.++### Skills (8)++| Skill | Purpose |+|---|---|+| `bb-methodology` | 5-phase non-linear hunting workflow + critical-thinking framework |+| `bug-bounty` | Master orchestrator |+| `triage-validation` | 7-Question Gate, 4 pre-submission gates, never-submit list |+| `report-writing` | H1 / Bugcrowd / Intigriti / Immunefi report templates, CVSS 3.1 + 4.0 |+| `security-arsenal` | Payloads, bypass tables, wordlists, gf patterns |+| `web2-recon` | Subdomain enumeration, host discovery, URL crawling |+| `web3-audit` | 10 DeFi bug classes, Foundry PoC template |+| `meme-coin-audit` | Token rug-pull detection |++### Slash commands (15)++`/hunt` `/recon` `/scope` `/triage` `/validate` `/report` `/autopilot` `/chain` `/intel` `/pickup` `/surface` `/remember` `/memory-gc` `/token-scan` `/web3-audit`**Repo**: https://github.com/shuvonsec/claude-bug-bounty**License**: MIT (verify in upstream repo)+---++## Tooling acknowledgments (not vendored — used as scaffolding)+### shuvonsec — `public-skills-builder`-Generator that produces per-class `hunt-*` skills from disclosed HackerOne reports. The 24 `hunt-*` skills bundled in `skills/` were generated by this tool.+Generator tool that produces skill scaffolding from disclosed HackerOne reports. Used to generate the initial scaffolds for the per-class `hunt-*` skills before the author's curation. The tool itself is not redistributed in this repo.**Repo**: https://github.com/shuvonsec/public-skills-builder**License**: MIT (verify in upstream repo)+---+## Inspirations### archangel / douglasday-A top-10 historical HackerOne hunter. The per-class `hunt-*` skill pattern with-specific chain templates from real disclosed reports was inspired by his-public skill stack screenshots:--- `hunt-rce`, `hunt-ssrf`, `hunt-xss`, `hunt-idor`, `hunt-rbac`, etc.-- `hunt <target>` shell command for engagement scaffolding-- Tool-integration skills (`ffuf`)+A top-10 historical HackerOne hunter. The per-class `hunt-*` pattern with chain templates from disclosed reports was inspired by his public skill stack screenshots, plus the `hunt <target>` engagement-scaffolding shell pattern.### Trail of Bits — `trailofbits/skills`-Reference for skill-authoring discipline. Their CLAUDE.md states:+Skill-authoring discipline reference. Their CLAUDE.md states:-> "Skills should be specific and actionable rather than reference dumps,-> focusing on behavioral guidance over comprehensive documentation."+> "Skills should be specific and actionable rather than reference dumps, focusing on behavioral guidance over comprehensive documentation."-This principle informed the `offensive-osint` refactor (lean SKILL.md +-`references/` subfolder).+This principle informed the `offensive-osint` v3 refactor (lean SKILL.md + `references/` subfolder for progressive disclosure).**Repo**: https://github.com/trailofbits/skills### SecSkills — `trilwu/secskills`-16 specialized security skills + 6 expert AI subagents. Demonstrated the-subagent pattern for complex multi-step tasks.+16 specialized security skills + 6 expert AI subagents. Demonstrated the subagent pattern for complex multi-step tasks.**Repo**: https://github.com/trilwu/secskills@@ -73,13 +130,13 @@ subagent pattern for complex multi-step tasks.- `travisvn/awesome-claude-skills` — general awesome-list- `VoltAgent/awesome-claude-code-subagents` — penetration-tester subagent pattern+---+## Tooling### PortSwigger — Burp Suite + MCP Server extension-Burp Suite Pro/Community is the foundation HTTP intercept tool. Their BApp-Store includes an "MCP Server" extension that exposes Burp's proxy history-to Claude Code via the Model Context Protocol.+Burp Suite Pro/Community is the foundation HTTP intercept tool. Their BApp Store includes an "MCP Server" extension that exposes Burp's proxy history to Claude Code via the Model Context Protocol.**Burp Suite**: https://portswigger.net/burp**MCP Server extension**: install via Burp's BApp Store@@ -94,38 +151,26 @@ The platform itself.### HackerOne API + Bugcrowd VRT-- HackerOne API enabled the `public-skills-builder` to pull disclosed reports-- Bugcrowd VRT (Vulnerability Rating Taxonomy) is referenced extensively in- `bugcrowd-reporting/SKILL.md`+- HackerOne's public disclosure program enabled the author to curate the per-class hunt skills from real-world report data.+- Bugcrowd VRT (Vulnerability Rating Taxonomy) is referenced extensively in `bugcrowd-reporting/SKILL.md`.-## Original contributions (this repo)--- `bugcrowd-reporting` skill — Bugcrowd VRT/OOS/severity tactics-- `evidence-hygiene` skill — cookie + PII + HAR redaction protocols-- `hunt` shell command — engagement-folder scaffolding-- Bundle packaging — single-step installer that vendors everything-- Documentation: `USAGE.md`, `INSTALL.md`, this credits file-- Architecture diagram (in USAGE.md)+---## Validation-Built and validated through a real Bugcrowd engagement on a major financial-target (May 2026). The engagement exposed four capability gaps that the-three original skills directly address:+Built and validated through a real Bugcrowd engagement on a major financial target (May 2026). The engagement exposed four capability gaps that the author's contributions directly address:++1. Hypothesis discipline (validation before drafting) — addressed by `triage-validation` (vendored)+2. Per-program reporting tactics — addressed by **`bugcrowd-reporting`** (original)+3. Engagement coordination / scaffolding — addressed by the **`hunt`** shell command (original)+4. Evidence hygiene / redaction — addressed by **`evidence-hygiene`** (original)-1. Hypothesis discipline (validation before drafting) — addressed by `triage-validation` from upstream-2. Per-program reporting tactics — addressed by **`bugcrowd-reporting`**-3. Engagement coordination / scaffolding — addressed by the **`hunt`** shell command-4. Evidence hygiene / redaction — addressed by **`evidence-hygiene`**+Engagement-specific identifiers (target names, account UIDs, endpoint names, bounty amounts) have been replaced with abstract placeholders in the shipped versions of `bugcrowd-reporting` and `evidence-hygiene`. Engagement details are not redistributed for confidentiality.-Engagement details are not redistributed in this repo for confidentiality.-The `bugcrowd-reporting` and `evidence-hygiene` skills shipped here are the-sanitized versions; engagement-specific identifiers (target names, account-UIDs, endpoint names, bounty amounts) have been replaced with abstract-placeholders.+---## License notes-- This repo's original work: MIT (see [LICENSE](../LICENSE))+- Original work in this repo: MIT (see [LICENSE](../LICENSE))- Vendored upstream skills retain their original licenses — typically MIT but verify each upstream source above-- If you're an upstream author and want attribution adjusted, removed, or expanded, open an issue or contact via the channels in your repo+- If you're an upstream author and want attribution adjusted, removed, or expanded, open an issue