SOTA Agent

SOTA Agent is a public ClawHub SOTA-campaign skill for CV and DS work. Use it when the user says "sota agent", "state of the art benchmark scouting", or want...

This skill appears coherent and appropriate for coordinating SOTA campaigns, but do these checks before installing or running it: 1) Inspect scripts/sota_public_safety.py (and the 4 omitted files) to confirm path/url sanitization and confirm there are no hidden network calls or credential usage. 2) Run the scripts in a dedicated campaign folder (not your home or /) so that file writes and any path resolutions stay isolated. 3) Do not pass secret-containing paths or API keys as arguments; the tool is designed to avoid requiring secrets. 4) If you plan to run any execution lanes (browser automation, VMs, Colab), prefer the paired execution skill the docs recommend and review its install/credential requirements separately. 5) If you need extra assurance, run the scripts in a sandbox or inspect them with a linter/static scanner to verify the omitted files contain no unexpected network or exec behavior.

Type: OpenClaw Skill Name: sota-agent Version: 1.4.2 The sota-agent skill bundle is a comprehensive framework for managing data science and computer vision research campaigns. It contains no evidence of malicious intent; rather, it includes significant security-conscious features, such as the 'sota_public_safety.py' utility and 'public-safety.md' documentation, which are specifically designed to redact secrets, API keys, and private filesystem paths from research artifacts. The bundle explicitly forbids the use of sensitive environment variables like OPENAI_API_KEY in favor of OAuth-backed sessions, and its scripts are limited to local file scaffolding and markdown rendering.