Supplementary recipe material, loaded on demand (through host.skills.read, for example) so that the main Skill stays progressively disclosed. These files write down contracts and design choices. They install nothing, and they supply no experimental validation.
| File | Responsibility |
|---|---|
admet.md |
How to install and invoke ADMET-AI, which direction each endpoint points in and how to aggregate them, what to expect at runtime, and what to do when it breaks. |
data_contracts.md |
The contracts the pipeline is held to: molecule identity, the columns every candidate and generation record must carry, the parent fields, the canonical operation_detail JSON, the lineage invariants, and what the visualization assumes about all of it. |
ga.md |
A starter GA to build from: structure, mutation and crossover options, chemistry filters, SA scoring, the composite score, and how to select for diversity. |