My agents are building a secure fork of OpenClaw
Show HN: SEKSBot – AI agents that can't see your secrets SEKSBot is a fork of OpenClaw where agents have zero access to API keys, tokens, or credentials — ever. The core insight is borrowed from SQL prepared statements: separate the instructions from the sensitive data. Agents write requests using named secret references. A broker intercepts and injects the real credentials at execution time. The agent never sees them. How it works: seksh (our nushell fork) has secure built-in commands (seksh-http, seksh-git) that route through the broker. Agents can make authenticated API calls and git…
What it does
In the maker’s words, at launch
Show HN: SEKSBot – AI agents that can't see your secrets SEKSBot is a fork of OpenClaw where agents have zero access to API keys, tokens, or credentials — ever. The core insight is borrowed from SQL prepared statements: separate the instructions from the sensitive data. Agents write requests using named secret references. A broker intercepts and injects the real credentials at execution time. The agent never sees them. How it works: seksh (our nushell fork) has secure built-in commands (seksh-http, seksh-git) that route through the broker. Agents can make authenticated API calls and git operations without the keys ever entering shell memory. seks-broker stores secrets and acts as a proxy. It can inject bearer tokens, API keys, and even handle asymmetric key signing — all without exposing anything to the agent process. Three layers of defense: (1) Agents never have secrets in env vars or memory. (2) The broker validates and scopes every request. (3) Skills use sandboxing on top of broker-mediated access. The problem we kept seeing: every AI agent framework puts API keys in environment variables. One prompt injection, one malicious webpage, one bad skill — and your keys are exfiltrated. We decided the only real fix is making it physically impossible for the agent to access them.
Does the same job
all alternatives →
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
