The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill is a legitimate tool for formal verification of reasoning claims.
[COMMAND_EXECUTION]: The skill provides CLI scripts (scripts/run_validate.py and scripts/run_formal_check.py) that perform validation tasks by reading user-supplied JSON inputs and executing local logic. This is appropriate for the skill's stated purpose.
[PROMPT_INJECTION]: The skill processes user-supplied reasoning and formal claims, which is a surface for indirect prompt injection. However, the implementation is assessed as safe due to documented security controls.
Ingestion points: Data enters through JSON payload files (--task-json, --claim-json, --step-json) and CLI flags in the bundled scripts.
Boundary markers: While explicit prompt delimiters are not shown, the process of formalizing natural language into logic expressions acts as a strict boundary.
Capability inventory: The skill executes local Python scripts and potentially calls external SMT/Lean binaries.
Sanitization: The references/formal-checking.md documentation specifies a strict safe subset of allowed constructors and prohibits imports or arbitrary code execution within formal expressions.

validate