The Scheduled Agents We Run, With the Prompts
How RunWhen uses RunWhen, part two: the scheduled Agents behind our vulnerability, certificate, PR review, environment health, storage and test-failure reports, with the prompts that drive them.
agents
6 posts with the tag “agents”
Find the leaks before you build the agents
Automating operations has no natural starting point, so most teams rank by intuition. We read a ticket estate instead. Three patterns show up again and again: hand-offs cost 5.6x, answering beats fixing four to one, and it is never one automation but thousands of small ones.
Safe by Design
Giving an agent production access is the easy part. Making it safe is a set of design decisions: a reviewed skill registry, RBAC through scoped assistants, read-only by default, human-in-the-loop, the Agent-Runner pattern, and a complete audit trail. 500,000+ production hours, zero AI-caused incidents.
Safe-for-Prod Harness: The Agent-Runner Pattern
Agents shouldn’t hold credentials. The Agent-Runner pattern separates the reasoning plane from the execution plane. Blast radius gets scoped per network boundary, and agent load on your infrastructure becomes a setting.
How RunWhen Uses RunWhen
A screenshots-first tour of how our own team — SRE, platform, developers, QA and the CEO — actually uses RunWhen Agents day to day. Daily newsletters instead of alerts, Slack root-cause answers, FinOps in minutes, and release readiness before standup.
New whitepaper: Safe For Prod — Building Operational Agents
A new whitepaper by Kyle Forster on why agents with freeform skills will never be safe for production, and why the durable investment is a library of narrow, scoped safe skills that every later agent reuses. Free PDF download.