Tap

[Tap @Tap_Mask]

给你的 AI Agent 装个安全带。

pip install traccia-guard

自动拦截：

3 行代码：
from traccia_guard import GuardSession
session = GuardSession("任务名")
session.record("shell_exec", {"command": "rm -rf /"}) # 自动拦截

GitHub: https://t.co/YHgjiCkJOd

[Tap @Tap_Mask]

#Aegis Agent OS
Trust infrastructure for AI agents. Not another framework -the layer that makes agents verifiable, accountable, and safe.
Request Policy Execution Proof Reputation Chain
8 layers, 14 Rust crates, 231 tests, zero warnings. Apache 2.0.
(https://t.co/MAgzlR7svw)

[Tap @Tap_Mask]

推文：

I let an AI agent trade for me. A hidden prompt injection told it to use 20x leverage.

AES caught it before liquidation.

pip install aes-mcp

🎥 Full demo ↓

[Tap @Tap_Mask]

"Prompt Injection almost liquidated my BTC trade"

[Tap @Tap_Mask]

A. 推广文章：为什么 AI 需要“行为证据”
标题：为什么我们不再相信 AI 的日志——以及一个正在解决这个问题的开放协议

过去一年，AI Agent 能做越来越多的事。它们能写代码、调用 API、���理日历、操作资金。但我们一直回避一个核心问题：

你如何证明一个 AI Agent 真的做了什么？

[Tap @Tap_Mask]

为什么ai需要行为证明

[Tap @Tap_Mask]

给 AI 调用加了个“行车记录仪”。
一行 base_url 改动，自动生成不可篡改的审计回执。
开源、Rust 编写、Docker 一键部署。
自带安全审计报告。

🔗 https://t.co/XVuxAFFeLz
#opensource #rust #ai #deepseek #audit

[Tap @Tap_Mask]

# Building a Microsecond-Level AI Agent Execution Engine with Rust

[Tap @Tap_Mask]

⚙️ agent-core: Rust-based AI agent execution engine
✅ Exactly-Once | 🔒 WASM sandbox | 📊 Prometheus metrics

🌐 Distributed Redis scheduler with retries & dead letter
⚡ 1.15ms Redis ops, 828ns WASM calls

Open-source: https://t.co/jjgcL3bhhF
#Rust #AI #Web4 #OpenSource

[Tap @Tap_Mask]

🚀 Introducing agent-core: a microsecond-level AI agent execution engine written in Rust.

✅ Exactly‑Once guarantees (Jepsen‑tested)
🔒 Secure WASM sandbox (memory/CPU limits)
📊 Built‑in Prometheus metrics

[Tap @Tap_Mask]

⚙️ agent-core: Rust-based AI agent execution engine
✅ Exactly-Once | 🔒 WASM sandbox | 📊 Prometheus metrics
🌐 Distributed Redis scheduler
Open-source | #Rust #AI #Web4
链接：填写 GitHub 仓库��址：https://t.co/jjgcL3bhhF

[Tap @Tap_Mask]

@ThisWeekInRust Show HN: agent-core – A modular Rust execution engine for AI agents with Exactly-Once guarantees

GitHub: https://t.co/jjgcL3bhhF
https://t.co/YakWJzFIWf: https://t.co/VcCYK9jgF2

Any feedback is welcome – especially from folks building production AI agents!

[Tap @Tap_Mask]

@rust_foundation
Show HN: agent-core – A modular Rust execution engine for AI agents with Exactly-Once guarantees
GitHub: https://t.co/jjgcL3bhhF
https://t.co/YakWJzFIWf: https://t.co/VcCYK9jgF2

Any feedback is welcome – especially from folks building production AI agents!

[Tap @Tap_Mask]

Looking for contributors! agent-core needs help with:
🔹 Distributed scheduler (splitting control/data plane)
🔹 WASM sandbox for extensions
🔹 Idempotency & retry policies
If you like Rust + distributed systems, come join the fun!
👉 https://t.co/FKiNZLVIgx
#opensource #rustlang

[Tap @Tap_Mask]

One design principle I’m proud of: all 19 modules have frozen interfaces.
No more top‑level modules will be added – new capabilities come via the extension slot (WASM sandbox planned).
This keeps the core small and auditable.
Thoughts? https://t.co/jjgcL3bhhF

[Tap @Tap_Mask]

🚀 I just open‑sourced agent-core – a modular Rust execution engine for AI agents.
It handles concurrency, retries, observability & persistence so your agents can actually do things reliably.
⚙️ Built with Tokio, DAG workflows, and 19 frozen modules.
👉 https://t.co/jjgcL3bhhF

[Tap @Tap_Mask]

我刚刚在欧易领水中心获得了 0.01 TESTNET_SAHARA！快加入获取更多测试网代币吧！ https://t.co/YoFWa1meQ1