BugReaper

Structured web2 bug bounty AI skill — 18 vulnerability classes, 4 bug bounty platforms, zero AI slop. Compatible with OpenClaw, Cursor, Claude Code, Antigravity, and Windsurf.

BugReaper is an Agent Skill that turns any compatible AI agent into a disciplined web2 bug bounty hunter. It enforces evidence-based validation, simulates real triage for HackerOne, Bugcrowd, Intigriti, and YesWeHack, and chains low-severity bugs into critical findings. Every finding requires a working PoC before it gets reported.

---

Agent Compatibility

The Agent Skills format became an open standard in December 2025. BugReaper installs into all compatible agents without modification.

---

What’s Inside

bug-reaper/
├── SKILL.md                     # Agent trigger + 4-phase workflow
├── references/
│   ├── recon.md                 # 7-phase recon methodology
│   ├── audit-rules.md           # Strict evidence requirements
│   ├── exploit-validation.md    # Input → sink tracing
│   ├── false-positive-elimination.md  # Adversarial FP checklist
│   ├── severity-guide.md        # CVSS scoring + platform tier map
│   ├── waf-bypass.md            # 15 WAF products, 10 bypass techniques
│   ├── chaining.md              # 8 chain templates (P3 → P1 escalation)
│   ├── platforms/               # HackerOne · Bugcrowd · Intigriti · YesWeHack
│   └── vulnerabilities/         # 18 hunting methodology files
└── scripts/
    ├── analyze_scope.py         # Parse program scope → structured JSON
    └── generate_report.py       # Generate platform-specific Markdown reports

18 vulnerability methodologies — each with confirmation payloads, bypass techniques, evidence requirements, and “do not report” rules that reflect real triage patterns.

---

Install

OpenClaw / ClawHub

/install bug-reaper

Or search for bug-reaper on ClawHub.

Cursor, Claude Code, Antigravity, Windsurf

# From your project root
git clone https://github.com/shaniidev/bug-reaper .cursor/skills/bug-reaper     # Cursor
git clone https://github.com/shaniidev/bug-reaper .claude/skills/bug-reaper     # Claude Code
git clone https://github.com/shaniidev/bug-reaper .agents/skills/bug-reaper     # Antigravity

The skill auto-triggers when you mention bug bounty, pentest, find vulnerabilities, or any vulnerability class name in your agent conversation.

---

How It Works

Phase 1 — RECON (references/recon.md) Passive subdomain enumeration, tech fingerprinting, JS bundle mining, endpoint discovery, attack surface mapping. Seven structured steps before touching a single payload.

Phase 2 — AUDIT (references/vulnerabilities/) 18 vulnerability classes ordered by bounty ROI. Reads the relevant methodology file for each class — confirmation payloads, bypass techniques, and what defenses to verify before claiming exploitability.

Phase 3 — VALIDATE (references/exploit-validation.md + references/false-positive-elimination.md) Traces attacker-controlled input from entry point to dangerous sink. Applies an adversarial checklist that actively tries to disprove each finding before it’s reported. Findings stay Theoretical until real PoC output is provided.

Phase 4 — REPORT (references/platforms/ + scripts/generate_report.py) Generates a platform-appropriate report. Triage checklist, severity scoring, and report template match the target platform’s actual acceptance criteria.

---

Vulnerability Coverage

Each file includes: detection probes · confirmation payloads · defense bypass techniques · evidence requirements · impact classification · “do not report” rules.

---

Scripts

Analyze a program scope file:

python scripts/analyze_scope.py hackerone_program.md --output scope.json

Generate a platform-specific vulnerability report:

python scripts/generate_report.py \
  --platform hackerone \
  --vuln-type idor \
  --input finding.json \
  --output report.md

Supported platforms: hackerone · bugcrowd · intigriti · yeswehack
Supported vuln types: xss · sqli · nosqli · ssrf · idor · auth · biz-logic · cors · csrf · rce · ssti · lfi · xxe · open-redirect · subdomain-takeover · prototype-pollution · http-smuggling · api-graphql

---

Sample Finding Output

When BugReaper surfaces a vulnerability, it uses this structure:

Title: IDOR on Order History — Any User's Orders Accessible
Severity: High
Confidence: Confirmed
Attack Prerequisites: Authenticated user (any account)
Vulnerable Endpoint: GET /api/v2/orders/{order_id}
Attack Path:
  1. Authenticate as User A, place an order → note order_id
  2. Authenticate as User B
  3. Request GET /api/v2/orders/
  4. Full order details returned — items, address, payment summary
Why This Is Exploitable: No ownership check on the orders endpoint. The
  backend retrieves the order by ID alone with no session validation.
Realistic Impact: Any authenticated user reads another user's full order
  history including shipping address and last 4 card digits.
PoC Request:
  GET /api/v2/orders/10482 HTTP/1.1
  Authorization: Bearer 
Suggested Verification: Run the above request. Confirm order 10482 belongs
  to a different account than the token.
Recommended Fix: Validate req.user.id === order.userId before returning.

---

Contributing

PRs welcome — additional platform support, new vuln classes, updated bypass techniques, or improved triage checklists. Open an issue first for significant changes.

---

License

MIT © 2026 shaniidev