Regression benchmark orchestrator
Skill alainlebret/claude-agents/higher-ed-teaching-agents/skills/regression-benchmark-orchestrator
Runs repeatable cross-workflow benchmarks to detect drift in grading stability, artifact quality, and policy compliance across model or prompt revisions.From its SKILL.md
npx -y skills add alainlebret/claude-agents --skill regression-benchmark-orchestratorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
0.9 KB, 138 tokens by cl100k_base, as published. Nobody here has run it
Regression Benchmark Orchestrator
Goal
Provide early warning when system changes degrade quality or consistency.
Inputs
- Baseline benchmark cases
- Candidate workflow outputs
- Historical benchmark reports
Produce
Write:
regression-benchmark-report.jsonregression-diff.md- optional
benchmark-traces/
Benchmark dimensions
- grade stability across repeated runs
- quality gate pass-rate deltas
- language-policy drift
- artifact completeness drift
- runtime and token-cost drift
Rules
- Compare against explicit baseline version tags.
- Flag statistically or operationally meaningful regressions.
- Recommend rollback or investigation when required thresholds are breached.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.