Havuz

白箱AGI架构探索：元认知（自我认知循环）、持续学习（知识飞轮）、世界模型（条件空间+语义时空图）、自我改进（自举纪律）、零LLM白箱管线与可审计信任护栏。

DeepSeek Harness gate - per-turn persistent-operation counter: remind, instruct the model to stop, then ask for confirmation. Steps aside un

Deterministic safety solutions for probabilistic AI agents

Agent Execution Partnership AEE is an open-source control plane that ensures every AI agent action is authorized before it runs, observable

The open standard for runtime agent control — declarative hooks, policy enforcement, and observability across AI agent frameworks.

Agent-ergonomic CLI for TypeSafe's Jev: fast calibrated judgments (pick, rate, check, rank, triage, guard) from the shell

An unattended maintainer for one repository. It reviews, files, prunes, and merges only against proof that a guard bites.

A measured tool-call firewall for AI agents: allow, ask or deny before the agent acts. Claude Code mod and hook, MCP proxy. Built on TypeSaf

Ekbasis: an open world model for agents. What an action will do, before it is done.

A really fast and lightweight harness for AI agents

An open taxonomy and scoring framework for evaluating AI agent sandboxes: 7 defense layers, 7 threat categories, 3 evaluation dimensions, 27

Fast, deterministic verifier and gate for AI agents: exact math, claim and citation checks, shell/write policy, hash-chained step ledger and

Local-first verification layer for AI agents: evidence, provenance, policy gates, human approval, and auditable Trust Receipts before a clai

Undo for AI agent shell commands. Snapshots files before Claude Code runs destructive bash (rm -rf, git reset, rsync), so agent mistakes are

Open-source runtime for grounded, model-agnostic computer-use agents with verifiable action receipts.

AI agents should not make irreversible changes. TraceFold escrows the inverse before an effect lands, or refuses it, and issues receipts any

Reversible side effects for AI coding agents — snapshot files, hold PostgreSQL writes, and review outbound actions before they become perman

The agent's prompt shows you the command; Termaxa shows you the consequence. Previews, backups, policy, an audit record. Claude Code, Codex,

Trust nothing. Ship safely. — Skeptical-reading and prompt-injection defense skill for AI agents. Provenance tagging, red-flag patterns, ref

Protect AI agents from runaway loops, repeated tool calls, and uncontrolled execution. Provider-independent, zero-dependency, TypeScript-fir

Open-source local MCP runtime that checks AI-agent tool calls, gates risky actions on approval, verifies results, and avoids blind retries a

OpenClaw-compatible MASL safety gate with public RAG packs for memory-aware AI agents

Runtime safety net for LLM agents. Detects token spirals, kills doomed tasks early, tells you exactly why. Rust core, Python SDK. pip instal