JulienBrouchier

6 ms·

1. 1.▲
   by JulienBrouchier 2mo ago

   This is (yet another) example of agentic computing being the great mixer. It mixes trusted and untrusted, private and public, code and data. This is the hard part when you build an agentic system : being able to compartmentalise your agents
2. 2.▲
   by JulienBrouchier 3mo ago

   It is a two step process : - First you create a container/sandbox/machine to contain your agent in it. If you are a dev, dev containers are a good solution. If you use OpenClaw, OpenShell is nice. there are plenty of examples on
3. 3.▲
   by JulienBrouchier 4mo ago

   There are positive things happening too. Recently, npm released min-release-age filtering the lets you only import packages that have a certain vintage, protecting you from all these supply chain attacks as they tend to be quickly detected.
4. 4.▲
   by JulienBrouchier 5mo ago

   Author here. I'm on the Trent team. This is a technical analysis of ClawHub's skills landscape. Happy to answer questions about the methodology or where you think we should look next (or again).
5. 5.▲
   Distinguishing Malicious from Vulnerable: 2,354 Popular ClawHub Skills Analysed(trent.ai)

   5 points by JulienBrouchier 5mo ago|1 comments