ShawnC21

5 ms·

1. 1.▲
   Show HN: ClawZero – execution boundary stops prompt injection exfil in OpenClaw(github.com)

   1 points by ShawnC21 7mo ago|0 comments
2. 2.▲
   by ShawnC21 7mo ago

   The prompt injection risk CNCERT is describing is structural.. it's not specific to OpenClaw. Any agent runtime where model output can reach privileged tools without an enforcement boundary has this exposure. The agencies' advice
3. 3.▲
   by ShawnC21 7mo ago

   I'm Building MVAR: It's a deterministic execution firewall for LLM agents. The core idea: prompt injection becomes dangerous only when model output can trigger privileged execution.. Instead of detecting malicious prompts, MVAR en
4. 4.▲
   by ShawnC21 7mo ago

   One clarification: MVAR is not a prompt filter and not a model judge. The enforcement happens at the execution boundary. If model output reaches a critical sink (shell, filesystem, credentials, etc.) with untrusted provenance, the runtime b
5. 5.▲
   MVAR: Deterministic execution firewall for LLM agents (50 attacks blocked)(github.com)

   1 points by ShawnC21 7mo ago|1 comments
6. 6.▲
   by ShawnC21 8mo ago

   Author here. MVAR is an IFC-based reference monitor for AI agent runtimes. Rather than attempting to detect prompt injection at the model layer, it enforces deterministic policy at privileged execution sinks. Core invariant: UNTRUSTED + CRI
7. 7.▲
   Show HN: MVAR – Deterministic sink enforcement for AI agent(github.com)

   1 points by ShawnC21 8mo ago|1 comments