hunt-business-logic is a skill for Claude covering business logic vulnerabilities: price and quantity tampering, verification flows enforced only in the UI, payment webhooks with no signature check, phone and callback verification trust, and internal pages exposed on production domains. It draws on twelve public bug bounty reports including coupon-race-stacking cases and price-field manipulation across e-commerce and payment platforms.
The skill covers attack surface signals in URLs, headers and JavaScript bundles, a step-by-step hunting method, payload patterns for header rotation and payment tampering, common root causes, and a defense-to-bypass table. A three-question gate (what the attacker can do, what the victim loses, and whether it reproduces in ten minutes) sits in front of every finding before it is written up. It is for assets covered by a program or engagement you are authorized to test.
What You Get
- Attack surface signals across URL patterns, response headers and JS bundles for checkout, verification and internal-tooling endpoints
- A ten-step hunting method covering authentication boundary mapping, verification-flow testing, rate-limit header rotation, and payment interception
- Payload patterns for header-rotation rate-limit bypass, in-flight payment tampering, unauthenticated internal-page discovery, and verification-gate bypass
- A defense-to-bypass table covering IP-based rate limiting, CAPTCHA, email verification gates, webhook signatures, and phone/OTP trust
- Twelve disclosed report citations from 2016 to 2023 covering coupon races, negative-quantity pricing, mass-assignment, and archived-price swaps
Common Use Cases
- Testing whether a checkout or subscription endpoint trusts a client-supplied price or quantity field
- Checking whether an email or phone verification gate is enforced in the API, not only in the frontend UI
- Searching JS bundles and sitemaps for internal or employee-only pages deployed without authentication on a production domain
What Makes It Unique
The skill keeps a financial-impact lens throughout: every disclosed case it cites demonstrates a measurable dollar loss or a quantifiable platform-wide exposure, and its validation gate asks directly what the victim loses, giving a logic flow an attributable dollar figure before it is scored. It also links each pattern forward into related classes such as race conditions, IDOR, mass assignment and account takeover, framing a logic bug as the first step of a larger chain.
Under The Hood
The skill is a single SKILL.md of guidance with no bundled scripts. Following it, the agent sends HTTP requests with rotated headers and tampered price or quantity fields to checkout, subscription and verification endpoints, inspects JS bundles and robots.txt/sitemap.xml for internal paths with grep-style patterns, and intercepts payment redirects to test server-side validation. Use it only against assets you are authorized to test.