waypoint-bio

SUMMARY

What it does

This skill provides procedural knowledge for working with Outpost Bio's open microbiome foundation models, including the Waypoint checkpoints (6M, 45M, 170M parameters), the Atlas pretraining corpus, and the Compass benchmark. It covers embedding microbiome samples, fine-tuning on taxonomic abundance data, benchmarking on Compass, pretraining a GPT-2 style model on taxonomic abundance profiles, and converting abundance tables from MetaPhlAn, Kraken2, QIIME 2, and MGnify into the waypoint format. The skill is based on the SKILL.md file from the K-Dense-AI/scientific-agent-skills repository, which was inspected and found to contain detailed instructions, including setup, data format specifications, workflow steps, scientific caveats, and references. The skill declares an MIT license and requires Python 3.10+, the waypoint-bio package, network access, and a Hugging Face token with access to gated outpost-bio repositories. A GPU is recommended for pretraining and benchmarking.