What problem does it solve? Security patches often fix only the reported payload or a single code path, leaving sibling endpoints, alternate encodings, or race windows exploitable. This Skill provides a structured method to determine whether a CVE fix is actually complete before trusting it. ## Core Features & Use Cases - Invariant Reconstruction: Rebuilds the original security invariant from the advisory and diff, then checks every reachable equivalent path against it. - Bypass Pattern Detection: Flags recurring patch-failure motifs such as incomplete blocklists, normalization-after-validation, TOCTOU races, unpatched sibling routes, and missing regression tests. - Verdict with Evidence Discipline: Returns Complete, Incomplete, or Suspicious with a three-label split (technically_vulnerable, in_scope, program_reportable) and a 12-category negative-control taxonomy. - Use Case: Given a CVE fix commit for a path traversal bug, analyze the diff, discover the fix only blocks ../ but not ..\ or double-encoded variants, and produce a controlled reproducer proving the bypass on an authorized target. ## Quick Start Ask the AI to review this CVE fix commit and determine whether the patch completely closes the vulnerability across all sibling code paths.
