SkillSpec is a Hoare-style framework that treats the correctness of reusable agent ‘skill’ abstractions as a specification-reasoning problem, targeting failure modes like intent conflicts that are subtler than conventional code defects and can be silently masked by the underlying model. It converts a heterogeneous skill repository into a unified graph aligning natural-language descriptions, instructions, and code artifacts, then derives a formal expected specification (ExpectSpec) for each skill node — grounding correctness checking in explicit task boundaries rather than ad hoc testing.
