A security hook for coding agents that uses Jev to risk-score tool calls, detect prompt injection, and check skills/plugins. Works with Claude Code, Codex, Copilot, Gemini, Cursor, pi, OpenCode, and ACP.
该仓库真实使用 TypeSafe AI 的 Jev 模型，实现了编码代理的安全钩子，调用 Jev API 进行工具调用风险评分、提示注入检测和技能/插件检查，并包含实测数据、CLI 及多种代理集成，属于实质性的开发工具与安全防护项目。
NyarlathotepppppNyarlathoteppppp/pi-heedRuntime constraints for the pi coding agent: checks every side-effecting tool call against user-stated rules, powered by TypeSafe Jev. Includes a detailed benchmark and experiment log evaluating Jev's System One decisions.
cwjechw98-langcwjechw98-lang/jev-gatesThree guardrail gates for coding agents: pre-action approval, evidence-based completion, and rubric regression checks, powered by TypeSafe Jev (System One). Includes CLI scripts, offline tests, and CI workflows.
InverxInverx/JevREJevRE is a Python CLI integrating TypeSafe AI's Jev System One model to recommend the next investigation step in Android reverse engineering. It includes a real Jev API client and an offline mock mode.
