Orchestrate measurement, hypothesis, and judge scoring loops across variants.
npx skills add https://github.com/kwam1na/athena --skill ce-optimize-kwam1na
Please help me install this Agent Skill. Skill: ce-optimize Source: https://github.com/kwam1na/athena/tree/main/.agents/skills/ce-optimize Command: npx skills add https://github.com/kwam1na/athena --skill ce-optimize-kwam1na
Coordinate end-to-end optimization loops by orchestrating measurements, hypotheses, and judge scoring across variants.
Start by defining an optimization spec in references/optimize-spec-schema.yaml and run the ce-optimize workflow to begin the first measurement baseline, followed by hypothesis backlog generation and batch experimentation.
High-intent search queries and answers about installing and using this skill.
Automating metric-driven optimization experiments involves orchestrating measurement harnesses, establishing baselines, and iterating through hypothesis testing. This process coordinates quantitative metrics and judge scoring across multiple variants while maintaining an auditable experiment log.
Running LLM-based judge scoring during software optimization requires stratified sampling and batch evaluation across variants. This approach combines hard quantitative metrics with automated judgments to evaluate hypotheses and maintain strict checkpointing.
Setting up an optimization loop requires defining an optimization spec to establish a measurement baseline. The workflow then generates a hypothesis backlog and executes batch experimentation with hard checkpoints and diff-based commits on an optimization branch.
Yes, running parallel execution for software engineering measurement harnesses is supported. The optimization workflow coordinates end-to-end loops by enforcing hard checkpoints and evaluating multiple variants either serially or in parallel.
An auditable experiment log is needed when running iterative software optimization experiments that require traceability. It maintains strict checkpointing and diff-based commits on the optimization branch to ensure every measurement and hypothesis is recorded.
Yes, the optimization workflow supports both hard quantitative metrics and optional LLM-based judge scoring. It orchestrates these evaluation methods across variants using stratified sampling and batch evaluation to validate hypotheses.
psd401Design, run, and interpret controlled experiments to validate software changes.
tbroadleyExtract samples, messages, events, and metrics from Inspect AI .eval logs.
xotongAutomate metric-driven iterative optimization with parallel experiments and gating rules.
deathbeamGuide six-phase debugging to isolate software defects and performance regressions.
