augustus
Agent skills for placing, building, evaluating and improving decision-model systems, with Jev as the default hosted example.
Overview
Augustus equips coding agents to find useful placements for decision models, build decision-driven systems, create evaluations and hill-climb them against real outcomes. It is for moments when a step could be code, an LLM call, a classifier or ranker, or a person, or when a confidence score needs a threshold, abstention or human fallback, or when a prompt needs an eval before optimizing.
Its working model is evidence, bounded judgment, explicit policy, checked action and observed outcome, drawing on decision theory, value of information and signal detection. TypeSafe Jev (Choice, Score, Noul) is the default hosted exemplar, but it also covers classical classifiers, encoders, open decision heads, rankers and vision scorers, and acknowledges that sometimes the best answer is a formula, checklist or no model. A companion augustus-train skill covers data assembly, fitting, export and evaluation of application-specific models. It is independent of TypeSafe and needs no key for design guidance.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Should this step be code, an LLM, a decision model, a classifier or a human, and at what threshold?(inferred) | — |
How it works
- Identify a candidate decision step
- Choose the model family by task and test against a baseline
- Set an explicit policy with thresholds and abstention
- Evaluate and hill-climb against observed outcomes
Strengths
- Method-focused and vendor-neutral
- No key needed for guidance
- Covers when not to use a model
Caveats
- Guidance-heavy; results depend on the agent applying it
- Fixture executions are not general Jev claims
Getting started
Install the augustus and augustus-train skills from the repository into your coding agent.At a glance
- Type
- Agent skill
- Status
- beta
- Runtime
- agent skills
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Coding agents: independent augustus and augustus-train skills for application-specific decision models, covering primitive/base-model/method selection, data assembly, fitting, export/reload, bounded improvement and independent evaluation, with TypeSafe Jev as the default hosted exemplar.
yibie/awesome-jevUnofficial augustus and augustus-train agent skills for application-specific decision models: primitive/base-model/method selection, data assembly, fitting, export/reload, bounded improvement and independent evaluation; TypeSafe Jev is the default hosted exemplar
AnotiaWang/awesome-decision-modelsIndependent augustus and augustus-train agent skills for application-specific decision models, covering primitive/base-model/method selection, data assembly, fitting, export/reload, bounded improvement and independent evaluation; TypeSafe Jev is the default hosted exemplar.
cobanov/awesome-jevIndependent augustus and augustus-train agent skills for application-specific decision models: primitive, base-model and method selection, data assembly, fitting, export/reload, bounded improvement and independent evaluation; TypeSafe Jev is the default hosted exemplar.
hellogumbo/awesome-jev
Frequently asked questions
Is it tied to Jev?
No, Jev is the default hosted example but other model families are covered.
Do I need a key?
Not for design guidance; only to call Jev or another provider.
What does augustus-train add?
Data assembly, fitting, export and evaluation of application-specific decision models.
Related projects
Eval-first MCP server for Jev: prototype questions, map them over many items, and measure accuracy and thresholds on labelled data.
MCP serverbetaTypeScriptAgent skill that tells you when you need an eval and how to build it, with an opt-in lane routing closed-label residue to Jev.
BenchmarkbetaPythonClaude Code and Codex skills powered by Jev, published with a calibration curve measured on 4,995 real agent decisions.
Project- SkillRanker★ 134
Rust CLI that ranks which agent skills fit the next step from live session context with Jev, can abstain, and hooks into Claude Code.
CLIbetaRust Claude Code plugin that sends judgment calls to Jev, acts at 0.8+ for yes or 0.2- for no, and gathers more evidence in between.
Agent skillbetaJavaScriptUnofficial Claude Code plugin that checks 'done' claims and small judgment calls with Jev, logs receipts, and can warn on unverified stops.
Pluginbeta