v0.2.0

Existing-system insertion: audit a codebase or PR for judgment-shaped holes before inventing mappings.

Pegged against typesafe-ai/skills v0.5.7 (65a39f3, 2026-09-12).

Added

• Boundary-audit card (references/boundary-audit.md): three-way split (exact / bounded judgment / generation), code-smell catalog, fit test, opportunity map, smallest-viable-boundary rule, control layer around a generative model, centralized policy + raw-judgment retention, red flags, completion questions
• Protocol branch: audit first on existing systems; per-action risk gates; keep questions and thresholds in one reviewable module
• Skill-description trigger terms for brittle parsers, prompt-to-JSON classifiers, and agent loops that are really bounded decisions

Install

claude plugin marketplace add 24601/Augustus
claude plugin install augustus@augustus

npx skills add 24601/Augustus --skill augustus

Full notes: CHANGELOG.md

v0.1.0

First public release of the design-judgment skill for TypeSafe Jev.

Companion to typesafe-ai/skills v0.5.7 (65a39f3, 2026-09-12). That skill owns integration contracts; Augustus owns where judgment belongs. Re-read live TypeSafe docs before treating the pin as current API behavior.

Install

claude plugin marketplace add 24601/Augustus
claude plugin install augustus@augustus

npx skills add 24601/Augustus --skill augustus

Added

• Skill protocol, decision-design card, and evidence labels
• Classical-method mappings (features/utility, selective decisions, circuits, rerank, beam search)
• Agent self-assessment, optimizer coupling (Ax, DSPy, ProgramAsWeights)
• Toolbox sweep, methods catalog, 11-position composition algebra, question-design diagnosis
• Validation gates and offline evaluator
• Launch-week evidence archive (187 repos) and ecosystem index

Full notes: CHANGELOG.md