The Agent Skills Directory

[SAFE]: The skill acts as a technical manual for implementing secure and robust graph synthesis pipelines. No malicious prompt injections, exfiltration attempts, or obfuscated payloads were identified in the instructions or example code.
[DYNAMIC_EXECUTION]: The structural-verification.md file contains Python code demonstrating how to translate graph structures into executable torch modules. This interpretation is strictly limited to a predefined operator whitelist and serves as a necessary component for numeric verification of synthetic candidates, posing no risk of arbitrary code execution.
[INDIRECT_PROMPT_INJECTION]: The skill addresses the processing of untrusted, generated graph data. It provides defensive patterns such as structural verification, canonicalization, and numeric proofs to validate candidates before they are accepted, effectively mitigating risks associated with untrusted data ingestion.

using-structure-synthesis