
Declaw Arena — Can You Break the Sandbox? | Declaw
Control an AI agent attempting to exfiltrate data from a sandboxed environment.
ShivamNayak11
The full gallery
Tech stack
28 projects

Control an AI agent attempting to exfiltrate data from a sandboxed environment.
ShivamNayak11

Open-source remote control for Claude Code and Codex on iOS, Android, and web.
@Ex3NDR · X
Few days ago Happy had a birthday. Tried to refresh one of the most neglected part of it: our Happy Website: also i built a new harness called “rig”. i started to use at work recently instead of codex/claude code and i liked how it does work. i want to share this with everyone. it is a codex and cause code reimplemented using Pi, but keeping native prompts and expanded a little for uniformity (like added workflows to codex yay!). when something new released that is worthy of porting - i ask Sol to mim for prompts or read sources directly and i can have a feature within an hour. why this is needed? i was constantly blocked by the needing of integrating a lot of models together, which each has their own harness or many doesn’t have any, but i didn’t want to go too far from native and don’t like the idea of extensions - i want experience to be the same everywhere. i think configuration > extensions. more soon

Scan your AI bills to find waste and prove savings with tamper-evident evidence.
@askrashidkhan · X

Transfer encrypted files between AI agents via MCP with expiring links.
marcushorndt · HN
I made YAFL – a E2EE file handoff for AI agents

Monitor your external attack surface, detect subdomain exposures, and map live CVEs automatically.
@tugayakin34 · X
ThreatPort is an all-in-one Cyber Threat Intelligence & Attack Surface Management platform

Secure vault for organizing important documents with family access control.
@DwivediNitil · X
Go physical to digital


Automated backup and end-to-end encrypted disaster recovery for network device configurations.
Willdom1986 · V2EX
多年前考过 CCNP 网络工程师,以前有一个很头疼的地方就是网络设备配置丢失引发的各种故障。后来越发觉得,网络设备的配置实际上是一个非常关键的数字资产。 为此,我利用周末时间开发了 XConf ( https://xconf.ai) 来解决这个问题。 产品主要就是想解决两个问题:**自动化的灾备高可用**,以及**端到端的安全**。 * **自动备份与灾备**:支持将配置定时备份到云端。如果网络断连了,探针会自动在本地离线队列里暂存配置,等连接恢复后自动重传。 * **端到端加密**:本地的 Go 探针是完全开源的,会自动在内存中对密码进行流式脱敏,并在上传前通过 AES-256-GCM 在本地加密。SaaS 后台拿到的只有密文,解密和版本比对( Diff )都完全在你的浏览器本地沙盒里通过 WebCrypto API 进行。 * **打磨版的配置对比 Diff**:我花了一些功夫打磨了浏览器端的配置版本对比功能,力求让日常比对的交互体验足够丝滑。 * **AI 安全基线审查**:系统集成了 AI 辅助的安全基线审查,可以一键检查和分析配置中的潜在风险。 因为目前完全是我利用业余时间(主要在周末)维护,所以确实没精力提供工作日的在线即时客服。大家使用过程中遇到任何 Bug ,欢迎随时去 GitHub 提 Issue ,我会在周末集中修复和跟进。 目前公测已开放,不需要邀请码即可注册,欢迎大家体验吐槽。 探针 agent 开源地址: https://github.com/willdom-lee/xconf-agent 官方网站: https://xconf.ai

Analyze code for security vulnerabilities and verify code safety.
@POONAMSING9999 · X
Technology is getting better everyday so the vibe coding but still people loosing money.🚀 Around 62% of vibecode is unsecure. I made VibeGuard that analysis code and check whether it is safe,ready to go,every error and what it can cause, many more...

Crop, redact, blur, and export screenshots privately without uploading to any server.
@noshkass · X
Built SAFEFRAME a private, browser-only screenshot sanitizer. Crop, blackout, blur, and export sensitive screenshots without uploading a single byte. Try it: Built with ChatGPT Sites. @chatgptapp #Giveaway

Backtest crypto trading strategies by describing them in plain English.
@torquant · X
building Torquant. you build and backtest trading/investment strategies using only natural languange

Verify an AI agent's decisions and test tampering with the verification records.
foh_quarters · HN
Verify what an AI agent did, then tamper with the record (no signup)