0127Skills
Augustus
Augustus and augustus-train: agent skills for designing systems around Jev’s judgments, from task-specific model selection to bounded improvement and independent evals.
24601/AugustusREADME ↗
# Augustus Find, build, evaluate, and improve systems with decision models. [](LICENSE) [](https://github.com/24601/Augustus/releases) [](https://github.com/24601/Augustus/actions/workflows/quality.yml) [](https://24601.github.io/Augustus/) Augustus equips agents to **find useful placements, build decision-driven systems, create evaluations, and hill-climb them against real outcomes**. Reach for it when a step could be code, an LLM call, a classifier or ranker, or a person; when a confidence score needs a threshold, abstention, or a human fallback; or when a prompt or program needs an eval before you optimize it. It draws on decision theory, value of information, multi-criteria analysis, signal detection, search/control, and formal-methods boundaries across software, business, organizations, research, and everyday decisions. [TypeSafe Jev](https://docs.typesafe.ai/) (Choice, Score, Noul) is the default hosted exemplar. The skill also covers classical classifiers, encoders, open decision heads, constrained readouts, rankers, and vision scorers. Choose the family by the task, then test it against the baseline. Sometimes the best result is a formula, a checklist, or no new model. The working model is: **evidence → bounded judgment → explicit policy → checked action → observed outcome** Augustus is independent of TypeSafe. The [official TypeSafe skill](https://github.com/typesafe-ai/skills) and current provider docs own API contracts; Augustus supplies composition, implementation, evaluation, and improvement methods. Named for Augustus De Morgan, mentor of William Stanley Jevons. [](https://youtu.be/Frm5sVlQNI4) **[Watch the 3:05 film](https://youtu.be/Frm5sVlQNI4)**: Jevons' logic piano, De Morgan's boundary, how Augustus places and tests a bounded judgment, and h