AI Resonance · 阅读最新日报 · AI 入门推荐

2026-10-10 · America/Los_Angeles · 社区动态 · #10

laya-guard: a local guardrail that judges every tool call your coding agent makes

laya-guard: A local guardrail for coding agents that checks every tool call I built laya-guard, a local guardrail that checks tool calls made by your coding agent and decides whether to allow, review, or block them. It runs entirely locally, needs no API key, and takes around 90 ms per command on a CPU. Regex rules catch obvious risks like rm -rf /, curl ... | sh, and leaked API keys. Simple commands like ls and npm test skip the model entirely. Everything else is evaluated by Laya, a ~322M local model. It supports Claude Code, Codex CLI, Cursor, Gemini CLI, Copilot CLI, and 6 other agents.…

laya-guard: a local guardrail that judges every tool call your coding agent makes

热度 45.3 / 100;排名与评分保留该期记录。