What problem does it solve? Code reviews often miss unhandled boundary conditions—null inputs, race conditions, missing transactions, enum branches left silently unhandled—leading to production bugs like double charges, data loss, and silent failures. This Skill systematically walks every branching path in a diff or file and reports only the unhandled edge cases. ## Core Features & Use Cases - Five-Dimension Analysis: Covers Scalability (N+1 queries, missing pagination), DataConsistency (race conditions, missing transactions, named-set generalization), MigrationIntegrity (EF migration mismatches), ErrorHandling (missing timeouts, swallowed exceptions), and SkillFORBIDDEN (platform-specific forbidden patterns). - Structured JSON Output: Returns findings as a JSON array with location, trigger condition, guard snippet, consequence, severity, and dimension—directly writable to a review_findings database schema. - Standalone or Pipeline Use: Works independently on any diff, file path, or function, with no isolation restrictions, and supports optional also_consider hints for extra analysis scope. - Use Case: Paste a pull request diff for a payment service and receive a JSON list of unhandled paths, such as a concurrent double-charge risk flagged CRITICAL with a suggested distributed-lock guard. ## Quick Start Ask the AI to run edge-case-hunter on a specific diff or file path and return the JSON findings array.
