What problem does it solve?
AI agents and MCP servers face "MCP Rug Pull" attacks where tool schemas are maliciously replaced after approval, causing agents to execute unintended or harmful operations. This Skill provides cryptographic schema verification using ECDSA P-256 signatures and TOFU key pinning to ensure schemas remain tamper-free from publication to execution.
Core Features & Use Cases
- Cryptographic Schema Signing: Sign tool schemas and skill folders with ECDSA P-256 keys and SHA-256 hashes, creating verifiable proof of integrity.
- Trust-On-First-Use Pinning: Automatically pin developer public keys on first verification and reject subsequent key changes, blocking key substitution attacks.
- Public Key Discovery: Locate developer public keys via the RFC 8615 .well-known discovery endpoint without manual configuration.
- Key Revocation: Check revoked keys through inline lists and standalone signed revocation documents with structured reasons.
- Offline Verification: Use trust bundles to verify schemas in air-gapped environments, CI/CD pipelines, and enterprise deployments.
- Skill Folder Signing: Sign entire AgentSkills directories (SKILL.md plus file manifests) to detect file tampering and ensure skill integrity.
- Cross-Language Support: Identical verification guarantees across Python, JavaScript, Go, and Rust implementations.
Quick Start
Use the schemapin skill to sign your AI tool schemas with ECDSA P-256 and verify their integrity on load to prevent MCP Rug Pull attacks.