The Agent Skills Directory

[SAFE]: The skill includes seven dedicated local auditing scripts (e.g., bed_validator.py, artifact_inspector.py) designed to validate genomic files and release artifacts for integrity and safety before they are processed by native code or the network.
[SAFE]: Instructions implement a strict 'Network and cache gate' that prohibits implicit downloads. The agent is required to obtain explicit user approval and perform allowlisting before calling network-capable functions such as from_pretrained or open_remote.
[SAFE]: Shared safety logic in scripts/_common.py enforces secure path handling, explicitly rejecting URI-like paths, home expansions, parent traversal, and symlinks in local validators.
[SAFE]: The skill proactively identifies genomic intervals and barcodes as identifying metadata and mandates redaction in reports and logs to prevent sensitive data exposure.
[SAFE]: All external package installations (Gtars via pip and Cargo) are pinned to specific, reviewed versions and include instructions for execution in isolated environments with resource limits.

gtars