The Agent Skills Directory

[SAFE]: The skill implements a static analysis pipeline that parses code into an Abstract Syntax Tree (AST) using the standard library ast module. This allows for structural validation of code without execution, mitigating risks associated with running untrusted content.
[SAFE]: The methodology explicitly instructs the agent to avoid executing the generated code to test parameters, citing security risks as the primary reason for this constraint.
[SAFE]: Introspection techniques (inspect, dir, getattr) are used strictly for metadata extraction (signatures, parameter names) from libraries already present in the environment to build a reference Knowledge Base.
[SAFE]: The skill mentions external package installation (e.g., pip install) only as a remediation step for missing libraries, which is a standard developer workflow targeting official registries.

detecting-correcting-hallucinations-llm-generated