Writing
Long-form essays on the technical and operational questions we encounter at Axiom — authorisation patterns for AI agents, PDPA-mapped evidence chains, deploying carefully into government contexts, and the shape of the work as it develops.
No. 01 · in draft
Authorising the agent: structured human-in-loop control for Claude-driven pentests.
Every action that traverses a real target requires a ticket. The ticket carries scope, authority, decay window, and a receipt the chain can be replayed against. This essay walks through the authorisation gate as we operate it — what it refuses, what it logs, what passes through, and how a Claude-orchestrated stack can run faster *because* of the gate, not slower.
In draft. Public release follows.
Essay metadata
- Topic
- Authorisation patterns
- Length
- ~3,200 words
- Audience
- Security architects
- Status
- draft · 2026
For an advance copy, contact hello@axiomblueprint.com.my.
In the meantime