Back to Blog
Blog series
3 posts · 29 min readStart here
If you have read nothing yet, read these in order.
1
May 9, 2026·15 min read
The Eight-Property Harness Audit: A 30-Minute Readiness Test for Production Agents
A field-ready audit for deciding whether an AI agent is protected by a real production harness. Score the eight properties, collect the evidence, expose the hidden gaps, and leave with a ranked fix queue instead of vague confidence.
Read
2
May 9, 2026·9 min read
AGENTS.md Done Right: The Navigation File That Actually Helps Coding Agents
AGENTS.md is no longer a cute repo convention. It is the instruction surface coding agents actually read. Here is how to make it short, scoped, testable, and useful: a navigation file, not a prompt dump.
Read
3
May 8, 2026·5 min read
End-to-End Refund: How 12 Primitives Compose in One Production Run
Twelve posts have built individual pieces of the harness. This one is the run that ties them together — a single ₹24,500 refund traced from invokeAgent envelope to a byte-equal replay five minutes later, with code at every transition.
Read