Skill UI

Autonomous experiment loop that optimizes any file by a measurable metric. The agent edits a target file, runs a fixed evaluation, keeps improvements (git commit), discards failures (git reset), and loops indefinitely. Ideal for optimizing code speed, reducing bundle size, improving test pass rates, enhancing prompts, or boosting content quality through iterative measured attempts.

This skill provides expert guidance on treating prompts as first-class infrastructure. It covers prompt registry design, systematic versioning, A/B testing, and building comprehensive evaluation pipelines. It ensures that prompt changes are rigorously tested and deployed safely, preventing quality regressions in production AI features at scale.

A comprehensive guide for preparing research artifacts (code, data, prompts, models) for EACL submissions. It details the two stages: an anonymized supplement for the initial review (ARR) and a fully licensed, public release for post-acceptance. Emphasis is placed on reproducibility, responsible NLP practices, and proper documentation.