benchmark plugins

by Lauri Gates

macOS-specific dev environment tooling — kitty session persistence, LaunchServices health, incident postmortems for kernel panics and WindowServer hangs, EndpointSecurity/EDR CPU diagnosis, long-uptime drift, and a repeatable threshold-scored performance benchmark suite

by tesslio

Measure and improve how well a plugin's skills help agents solve real tasks. Runs activation evals (does the skill self-activate?) and content evals (does it help solve the task?), diagnoses gaps, fixes instructions, and re-runs until scores improve.

by Jacky Wang

评估器合集——对任务产出和 Agent Skill 设计做证据化、多维度质量评估

by Paulo Soares

Review, refactor and benchmark Agent Skills across Claude Code, Codex, Hermes Agent and OpenCode, with runtime and model discovery and objective headless A/B benchmarks.

by Virto Commerce

Three-layer performance loop for Virto Commerce — BenchmarkDotNet A/B (L1) → k6 load harness (L2) → dotnet-trace allocation/CPU/DB attribution (L3), a machine-readable perf-verdict, and optimization PRs. Depends on vc-fix: it REUSES vc-fix's onboarding (/project-init, project-profile.json), fix routing (qa-fix-routing), the backend developer/reviewer agents, and dotnet-fix — and ships ONLY the perf-specific measurement delta (no duplication of vc-fix machinery). Advisory only, never a CI gate.

by yukineko

External-benchmark (SWE-bench Verified) runner for Claude Code plugins: a typed SWE-bench Instance model, a deterministic fixture-based JSONL loader, and a gated `download` subcommand that shells out to `curl` (no HTTP crate) to fetch the Princeton NLP dataset only on explicit invocation. Subscription-native (one bundled Rust binary).