What problem does it solve?
Fixing and evolving citation styles reliably is slow and error-prone because upgrades, migrations, and new style creation must preserve fidelity while following the project’s validation gates.
Core Features & Use Cases
- One human-facing command for all style work: routes upgrade, migration, or creation to the correct internal workflow without exposing internal skills to users.
- End-to-end pipeline with guardrails: runs autonomous steps with shared gates and explicit stop conditions, then outputs a complete deliverables contract (fidelity metrics, SQI delta, authority basis, divergences, QA verdict, and more).
- Quality and evidence discipline: classifies failures using the shared decision rules, checks divergence registers, and ensures QA passes before completion.
Example: when a researcher needs an American Medical Association style updated to the latest Citum model and format, run a single evolve command and get fidelity/SQI outcomes plus QA verdict and divergence notes.
Quick Start
Ask the AI to run style-evolve in upgrade mode for a specific style file path, then return the full contract results and routed sub-skill header.