hunt-api-misconfig

Send privileged fields such as is_admin, role, verified, or tenant_id in the body of profile, account, signup, or reset endpoints and check whether the server applies them. If the response or subsequent session reflects the elevated attribute, the endpoint blindly binds client input to the model.

Target JSON object-update endpoints like profile, settings, or preferences and inject __proto__ or constructor.prototype markers. Then trigger a separate sink endpoint and compare against baseline; changed defaults, errors, or job behavior confirm the polluted prototype reached a sink.

It occurs when user input is interpolated into a backend API path or query, such as /api/users/<username>. Sending metacharacters like %23, %2f, or ../ lets you traverse routes and append fragments like /field/passwordResetToken to read other users' sensitive fields.

It documents alg=none and RS256-to-HS256 confusion examples, but the frontmatter states JWT signature and crypto forging is owned by the hunt-jwt-crypto skill. This skill covers only non-crypto JWT handling, so use the dedicated skill for full JWT attacks.

OData operators like startswith, substringof, and eq look unlike SQL keywords, so SQLi blacklist signatures never fire. Encoded variants like %24filter or double-encoded %2524filter, plus $batch multipart wrapping, further evade WAFs that only scan canonical outer requests.

Stop at a controlled, non-destructive marker such as a polluted status code or JSON spacing property in production. Only read secrets or environment data when the scope explicitly authorizes it, and prove impact through admin job or rendering endpoints that consume polluted defaults.