SECURING AI WITH IDENTITY
I write about agent identity, authorization, secure harnesses, AI workflows, models, and the controls enterprises need to trust autonomous systems.
[HUMAN] -- delegates --> [AGENT]
| |
+---- [IDENTITY] -------+
|
[AUTHORIZATION]
|
[TOOLS] [DATA] [MODELS]
|
[AUDIT_LOG]- subdirectory_arrow_right
Agent Identity & Authorization
Designing how agents authenticate, delegate access, and stay inside clear permission boundaries.
- subdirectory_arrow_right
Harness & Workflow Security
Securing the orchestration layer that connects prompts, models, tools, memory, and enterprise data.
# SESSION EVIDENCE
Building an IAM Agent: Start With the Data, Not the LLM
How I built an IAM agent by treating the data and analytics layer as the source of truth, with skills and LLMs acting as a controlled interface.
Your Prompt Isn't the Security Boundary
A practical way to think about harness safety when agents can actually execute.
WHAT I WRITE ABOUT
This is where I document the security problems that show up when agentic systems move from demos into real enterprise environments. My focus is identity: how agents authenticate, receive authority, call tools, access data, and leave enough evidence behind for someone to understand what happened.
HOW I USE AI
I use AI as a research assistant, technical sounding board, and editor. The ideas come from my own engineering work, prototypes, and research. I challenge the output, verify technical claims, rewrite heavily, and take responsibility for everything published here.
ONE ARTICLE A WEEK
For the next year, I am publishing one practical article every week. Expect tool reviews, techniques, incident breakdowns, architecture notes, and lessons from things I am actively building. I will include the failure modes and trade-offs, not just the polished diagram at the end.