GitHub’s trending repos this month put agents to work rather than shipping new models. Strix runs autonomous AI penetration-testing agents that execute your app’s code, find vulnerabilities, validate each with a working exploit, and fix them in CI/CD. Anthropic-Cybersecurity-Skills packages 818 structured security skills for agents, mapped to six frameworks from MITRE ATT&CK to the newer Fight Fraud standard. oMLX serves models on Apple Silicon with continuous batching and SSD caching from the macOS menu bar, while llmfit scores models by whether they will run well on your hardware. career-ops turns any coding CLI into a job-search pipeline that scores listings, tailors CVs, and tracks applications. Cordis is the plugin framework underneath DeepSeek Harness, the open-source agent harness where everything is a plugin.
Agents now need skills and local compute, not just bigger models. Security shows it best: a developer can run a penetration test in hours instead of weeks, and a security team can hand an agent the same playbooks a senior analyst uses, so vetting skills matters more than ever. oMLX and llmfit make models fast enough to run on hardware you already own, keeping data and cost local.
The marketplaces NVIDIA’s SkillSpector scans for malware are now filling with security skills and agents, and with Cordis, teams can build their own agent harness instead of depending on one vendor’s product.
Read More: The same agent-tooling pattern appeared in an earlier August snapshot
Sources:
- Strix: Open-Source AI Penetration Testing (GitHub)
- Anthropic-Cybersecurity-Skills: 818 Security Skills for AI Agents (GitHub)
- oMLX: LLM Inference for Apple Silicon (GitHub)
- llmfit: Right-Size Models to Your Hardware (GitHub)
- career-ops: AI Job-Search Pipeline (GitHub)
- Cordis: A Meta-Framework of Spatiotemporal Composability (GitHub)
- DeepSeek Harness: Everything Is a Plugin (GitHub)
Disclaimer: For information only. Accuracy or completeness not guaranteed. Illegal use prohibited. Not professional advice or solicitation. Read more: /terms-of-service
Reuse
Citation
@misc{kabui2026,
author = {{Kabui, Charles}},
title = {GitHub {Trending} in {August:} {Agents} {That} {Test}
{Security,} {Run} {Locally,} and {Hunt} {Jobs}},
date = {2026-08-25},
url = {https://toknow.ai/posts/github-trending-ai-pentesting-security-skills-local-inference/},
langid = {en-GB}
}