Why the sandbox 'straightjacket' approach to securing AI ops agents is fundamentally flawed, and what a 'playground' paradigm—autonomy within bounded infrastructure, vetted workflows, oversight, and greylisted boundaries—looks like instead.
Randy Bias
A proof of concept, Nightcrier, for an automated triage system that receives events via MCP and general-purpose AI agents to investigate Kubernetes failures in near real-time using a combination of Agent Skills, MCP servers, and native APIs.
Randy Bias
A quick look at why today's general purpose agents like Claude and Codex are over-indexed on code development and how to change it.
Randy Bias
Anthropic releases Agent Skills as an open standard: turning general-purpose agents into AIOps and Kubernetes triage assistants with repeatable, safer workflows.
Randy Bias
Using Claude Skills and an operator-in-the-loop pattern to triage production Kubernetes outages with general-purpose agents
Randy Bias
