What problem does it solve? Rust APIs often leak implementation details—backend types, lifetimes, lock guards, Serde schemas, or runtime handles—through boundaries that claim to hide them, forcing consumers to couple to internals. This Skill audits Rust abstraction boundaries and reports only evidence-backed leaks, filtering out legitimate concrete contracts. ## Core Features & Use Cases - Five-condition confirmation rule: Every finding must prove the boundary, the intrusion, the consequence, source evidence, and the absence of an intentional contract before being reported. - Rust-specific leak taxonomy: Covers representation, backend/dependency, ownership/lifetime, concurrency/runtime, error, schema, invariant, generic, macro, unsafe/FFI, feature, and behavioral leakage categories. - Optional candidate collector: A dependency-free Python script inventories syntactic leak signals (public fields, foreign errors, escape hatches) as leads for manual semantic validation. - Use Case: Ask the agent to review a Rust PR for new leaky abstractions; it traces boundaries outward and consumers inward, then returns severity-ranked findings with file locations, change tests, and minimal remediation steps. ## Quick Start Ask the agent to abstraction-police your Rust workspace and report only evidence-backed leaky abstractions with file locations and severity rankings.
