I spent 6 months building a runtime governance layer for AI agents — here's what survived testing" (opens in new tab)

Agents are moving from demos to touching money, infrastructure, and customer data. Sentinel SCA is a runtime admissibility layer. Before an agent action executes, Sentinel evaluates the request and returns one of three verdicts: ALLOW REVIEW DENY Every decision is cryptographically signed and recorded in a tamper-evident audit ledger. This post is the validation report: what I built, what I tested, what passed, and what remains unproven. At the time of writing, Sentinel has been validated acr...