hunt-ato is a skill for Claude that organizes account takeover testing into nine paths. They include password reset flaws (host-header handling, predictable or reusable tokens, token leaks through Referer), email change without re-authentication, OAuth account linking, session handling, JWT weaknesses, recovery questions and chained cases such as a dangling subdomain used with OAuth redirects.
The skill treats takeover as a destination reached by several primitives. Each path becomes a reportable finding only when it is shown against a second test account from a separate session, and the skill lists checks that rule out false positives, such as reading the delivered email to see which link domain the server really sends. It is for assets in scope of a program or engagement.
What You Get
- Nine takeover paths with the observable signs of each
- A proof standard based on two test accounts and out-of-band confirmation
- False-positive checks such as reading the delivered email and capturing outbound requests in the network tab
- Notes on JWT, recovery-question and OAuth redirect weaknesses
- Chain ideas that connect takeover with subdomain and open-redirect findings
Common Use Cases
- Reviewing a password reset flow on an in-scope application
- Checking that email and password changes ask for current credentials
- Judging whether a low-severity redirect or token leak can be shown as a full takeover
What Makes It Unique
It treats account takeover as an outcome with many routes and ties each route to a proof requirement, which reduces reports that rest on reflected headers alone. It keeps the routes together in one skill so one test session can cover reset, session, OAuth and token paths.
Under The Hood
The skill is a single SKILL.md of guidance with no bundled scripts. Following it, the agent sends HTTP requests to reset, login and recovery endpoints of the target in scope, may set up out-of-band callbacks to confirm delivery, and may repeat requests to check rate limits on codes and recovery answers. It may write notes to the working folder. The text includes request snippets, token-handling examples and brute-force patterns as teaching material. Use it only with accounts and applications you are authorized to test.