ClawBench vs Scientific Agent Skills

ClawBench is a browser agent evaluation framework built around 153 real-world tasks running on 144 live production websites — not simulated environments or curated sandboxes. Tasks span e-commerce, travel booking, SaaS dashboards, government portals, and developer tools. A built-in request interceptor blocks genuinely irreversible actions (payments, form submissions that send data) so evaluations can run safely on real sites. The benchmark records five layers of data per run: session replays, screenshots at each decision point, raw HTTP traffic, agent reasoning traces, and browser action sequences. This makes failure analysis tractable — you can see exactly which DOM element the agent misidentified, not just a final score. The dataset is open and the evaluation harness is reproducible. The headline finding is sobering: Claude Sonnet 4.6, the best performer, completes only 33.3% of tasks. GLM-5 is second at 24.2%. No model exceeds 50% on any individual task category. The implication is stark — current browser agents are far from autonomous on the open web, and the gap between benchmark performance and production performance is still enormous.

Scientific Agent Skills is an open-source toolkit of 134 ready-to-use scientific domain skills for AI agents, covering cancer genomics, drug-target binding prediction, molecular dynamics, RNA velocity analysis, geospatial science, and time series forecasting. Each skill integrates with 78+ scientific databases and is backed by 70+ optimized Python packages, installable with a single npx command into agents like Claude Code, Cursor, or Codex. The core idea is separating scientific compute from the agent's reasoning loop. Instead of asking an LLM to hallucinate bioinformatics pipelines, you give it callable skills that actually connect to NCBI, PDB, ChEMBL, and other authoritative data sources. Optional cloud compute via Modal handles GPU-intensive workloads — molecular dynamics simulations, protein structure inference — without requiring local hardware. Forty-plus model integrations mean the skills layer is agent-agnostic. With 18.1k GitHub stars, this project is filling an obvious gap: the agent ecosystem has exploded in developer tools but scientific workflows have lagged behind. A bioinformatician can now wire up a Claude Code agent that genuinely queries gene expression databases, runs differential analysis, and interprets results — without writing custom integration code for each data source.