lean
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands via the Lean 4 build tool (lake) and a local bash script (lean_trust_audit.sh) to perform theorem proving and project audits.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Lean source files and configuration data, creating an attack surface for indirect prompt injection.
- Ingestion points: Lean source files (.lean), lake-manifest.json, and project metadata.
- Boundary markers: Explicit 'Verification boundary card' and trust assumption reporting are required for verification tasks.
- Capability inventory: Shell command execution and Lean code elaboration/execution.
- Sanitization: Relies on lexical auditing of source files for specific keywords.
- [DYNAMIC_EXECUTION]: The skill uses 'lake env lean' to elaborate and execute Lean source code, which may include user-defined tactics and macros.
- [EXTERNAL_DOWNLOADS]: The 'lake' build tool is used to manage and download project dependencies from remote repositories.
Audit Metadata