Execution · YYLO Skills · stable 2.1.1

benchmark-yylo

Prepare reviewed cases and frozen checklists, compare models and harnesses, and independently evaluate retained outputs with YYLO Benchmark's thin trusted-host runner.

Goal

Help the agent act with the right procedure.

Prepares reviewed cases and frozen project/task checklists; compares models and harnesses; independently evaluates retained outputs. Loss is computed by the runner, not invented by the judge. Trusted-host hygiene is not a sandbox, and paid execution requires explicit authority.

Expected request: [case or study goal] [models/harnesses] [constraints]

Install explicitly

Choose one skill or a version-pinned package installer.

The direct open-standard command selects only this slug from the repository default branch. YYLO complete-set installer: Stable 0.2.11. Standalone Ledger has its own installer selection; inspect its help before choosing a tag. Use commands only with the labeled version.

Available now · one skill
npx skills add yylo-dev/yylo-skills --skill benchmark-yylo
All YYLO skills · Stable 0.2.11
yy skills install --version 2.1.1

Canonical instructions

Read before you delegate.

Read the checklist runtime contract and review the canonical checklist guide and examples. The runner does not impose a mandatory pilot ceremony; a study protocol may do so.

The complete procedure is intentionally not duplicated here. Review its versioned SKILL.md and supporting files in the canonical repository.

Open source for benchmark-yylo →

← Return to all Skills