The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute standard Lean toolchain commands, including lake build, lake clean, and lake env lean. These are required for proof development and verification and are executed in the context of the Lean project environment.
[DYNAMIC_EXECUTION]: The workflow involves generating and compiling Lean source code to produce formally verified proof terms. Safeguards are in place, including instructions for auditing the exported declaration's type and transitive axioms to ensure the integrity of the results.
[INDIRECT_PROMPT_INJECTION]: The skill processes external mathematical statements and Lean projects. It mitigates risks via an "adversarial review" process involving multi-model verification and back-translation of formal definitions to ensure alignment with user requirements.
Ingestion points: Lean project files and mathematical statements provided by the user.
Boundary markers: Semantic checks via cross-family model reviews and mandatory back-translation of Lean definitions into natural language.
Capability inventory: Access to Bash, Write, Edit, and Agent tools for formalization and parallel review work.
Sanitization: Relies on the formal Lean kernel for logical verification and multi-model audits for semantic consistency.
[EXTERNAL_DOWNLOADS]: The skill references standard dependency management via lake exe cache get for fetching library caches, which is a common and necessary practice for projects using Mathlib.

lean-formalize