Two of today’s entries are about trusting code that an AI helped write. alibaba/open-code-review runs deterministic checks first and then places every AI suggestion on the exact line it refers to, which is what you want when a large diff has to be reviewed one comment at a time. cloudflare/security-audit-skill works the other side of the same problem: it maps a codebase’s architecture and trust boundaries, sends isolated hunter agents after the uncovered areas, and hands every candidate to a fresh verifier that tries to disprove it before anything reaches the report.

The rest of the list is a tour of what coding-agent skills now cover. hypit-ai/hypit rebuilds a reference video as a word-timed production workflow, latent-spaces/brag packages the project you just finished into a short launch video with ready-to-paste share copy, and Tencent/BrowserSkill lets an agent borrow tabs in the browser you are already logged into while you keep working. affaan-m/ECC adds reusable routines for planning, tests and memory, and deepseek-ai/deepseek-harness is a plugin-based agent runtime worth experimenting with.

Not everything today needs an agent. eternity4719/HowToLiveBetter is a Chinese guide where each life suggestion records what you spend, what you get back and how strong the evidence is, citing only journal papers and official documents. bilawalsidhu/gods-eye-view follows live aircraft and ships across an interactive 3D globe, and tt-a1i/archify turns an agent’s description of a system into architecture diagrams you can compare between versions.