ai-paper-reproduction

No overt malware or obfuscation is visible in this module, but it is a high-risk supply-chain execution orchestrator. When enabled, it can execute commands extracted from an untrusted target repository’s README (directly or via a training runner). It also reads config files based on README-derived hints and writes captured execution evidence to disk. This should be treated as an execution-on-untrusted-input design requiring sandboxing, strict command validation/allowlisting, and strong isolation beyond what is shown in this file.