The Agent Skills Directory

[SAFE]: The skill provides architectural guidelines and Python code for synthesizing and verifying structured graphs, such as neural network topologies. The logic is entirely local and does not involve credential access, external data exfiltration, or persistence mechanisms.
[SAFE]: The implementation patterns for dynamic neural network assembly in structural-verification.md utilize a strict operator whitelist to restrict execution to known and validated operations, mitigating risks of arbitrary code execution during graph interpretation.
[SAFE]: The skill addresses indirect prompt injection risks by establishing a mandatory verification gate that validates generator-emitted topology against a typed grammar and arity contracts before any downstream processing occurs.

using-structure-synthesis