Add yylo-benchmark registry card

[InsightFactoryAPP]

Summary

• Tool Name: yylo-benchmark
• Binary: yylo-benchmark
• Primary Category: test (also: agent)
• Homepage / Repo: https://github.com/yylo-dev/yylo-benchmark

YYLO Benchmark is a thin trusted-host experiment runner for agent tasks: independent attempts over historical Ledger tasks, supplied coding prompts, and workflows. Retained outputs can be evaluated later with different checks or judges; no combined score, winner, or capability inference is made. Categorized test + agent following the inspect-ai / swe-agent / deepeval entries. Sibling of the yylo CLI card (#9, merged); one tool per PR.

Contribution Rules & Checklist

• No manual README edits: only data/tools/yylo-benchmark.yaml added; README regenerated locally for validation only and not included in the diff
• Exact file naming: data/tools/yylo-benchmark.yaml, name: yylo-benchmark
• Allowed effects: execute_code, network_access, read_files, write_files, secret_exposure — all from the whitelist
• Guardrails: risk.level: high with two concrete guardrails
• Local validation passed:

  python scripts/lint_registry.py   → Validated 461 tool cards.
  python scripts/generate_readme.py → diff contains only the new yylo-benchmark row + count badges
  

Disclosure

Self-submission by the YYLO team (affiliated), prepared with AI-agent assistance. Every summary / use_when / avoid_when phrase is quotable from the live README. Happy to adjust category, risk framing, or wording to fit the registry's conventions.