The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides educational code templates for late chunking that use the trust_remote_code=True parameter to load specialized embedding models from Hugging Face. This is documented neutrally as it pertains to standard model-loading procedures for established vendors like Jina AI.\n- [EXTERNAL_DOWNLOADS]: Several code snippets include instructions to fetch necessary resources from well-known repositories, such as NLTK tokenizers and pre-trained weights for BGE or E5 embedding models from official sources.\n- [SAFE]: The skill consistently uses descriptive placeholders like 'your-api-key' for sensitive credentials in all provided code examples, promoting secure secret management.\n- [SAFE]: No evidence of prompt injection, data exfiltration, obfuscation, or malicious persistence mechanisms was found across the architecture and reference files.

rag-architect