Zero-knowledge proofs have a vast applicability in the domain of cryptography, stemming from the fact that they can be used to force potentially malicious parties to abide by the rules of a protocol, without forcing them to reveal their secrets. $\Sigma$-protocols are a class of zero-knowledge proofs that can be implemented efficiently and that suffice for a great variety of practical applications. This paper presents a first machine-checked formalization of a comprehensive theory of Sigma-protocols. The development includes basic definitions, relations between different security properties that appear in the literature, and general composability theorems. We show its usefulness by formalizing---and proving the security---of concrete instances of several well-known protocols. The formalization builds on CertiCrypt, a framework that provides support to reason about cryptographic systems in the Coq proof assistant, and that has been previously used to formalize security proofs of encryption and signature schemes