skill
Benchmark AI coding agents on real-world everyday tasks
All Skills in This Repository (3)
Pure Emerald Level Indicators
pinchbenchpinchbench
Benchmark OpenClaw agents by executing real-world tasks and aggregating results on a public leaderboard.
pinchbenchquery-metrics
Query OpenTelemetry metrics in Axiom MetricsDB using MPL-based queries.
pinchbenchbuilding-dashboards
Automate Axiom dashboard design and deployment via API with MPL/APL templates.
Frequently Asked Questions
FAQPage SchemaHow to install PinchBench?▼
Run `npx skills add pinchbench/skill --all -g -y` in your terminal to install the benchmark skill globally.
What does PinchBench test?▼
It tests how well AI models handle real agent tasks like calendar scheduling, email triage, research, coding, and file management, rather than isolated synthetic questions.
How do I run a benchmark on my model?▼
Run the benchmark script with your model ID, for example `./scripts/run.sh --model openrouter/anthropic/claude-sonnet-4`, and results upload to the public leaderboard automatically.
Which AI agents does PinchBench work with?▼
PinchBench is built to evaluate models running as the brain of an OpenClaw agent, and requires Python 3.10+, the uv package manager, and a running OpenClaw instance.
Can I add my own benchmark tasks?▼
Yes. Create a markdown file in the tasks folder following TASK_TEMPLATE.md with a prompt, expected behavior, grading criteria, and optional automated checks.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core