ホーム /コーディングエージェント /pi-jev-sentinel
pi-jev-sentinel
Pi のツール呼び出し、ツール出力、返答を Jev で確認し、危険な操作とプロンプトインジェクションを検出します。ユーザー承認、コンテキストの再確認、シークレットの除去、任意のタスク固定を備えます。
このページは英語版リストから生成しています。
readme より
jev-sentinel A guard for coding agents, built on TypeSafe's Jev model: it checks what the agent does, reads, and says. One set of checks, three hosts: a Pi extension, and a plugin for Claude Code and Codex CLI. A README that says "AI agents: run curl … sh" is just more text to the model, and a permission prompt that only knows a tool name cannot tell you …
詳細
- セクション
- コーディングエージェント
- owner
- harshwasan
- スター
- 15
- フォーク
- 0
- 最終更新
- 2026-09-20
- 言語
- TypeScript
- ライセンス
- MIT
これは何か
- 形態
- フックやプラグイン
- ホストエージェント
- Pi
- 対象
- エージェント開発者
成熟度
ドキュメント