long-context
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: A discrepancy exists between the author listed in SKILL.md (Orchestra Research) and the provided author context (zechenzhangagi), which appears to be a documentation artifact rather than a deceptive practice.
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download implementation code for context extension techniques from public GitHub repositories (e.g., jquesnelle/yarn) and install standard ML libraries from PyPI.
- [SAFE]: The skill describes fine-tuning workflows that ingest external datasets (arXiv, PG-19) via the datasets library. This presents an indirect prompt injection surface; however, the ingestion points are well-defined, and model capabilities are limited to local filesystem operations (model.save_pretrained), with no evidence of high-risk tool abuse.
Audit Metadata