Agent Fleet Supervisor Daily-Ops Kit

Run a real supervisor shift over a fleet of AI agents: paste last night's run export and get a triage board that catches silent green failures, a capacity-bounded exception queue with containment moves, drift and escalation calls against explicit thresholds, and a handover note whose carry-forward block starts tomorrow's shift.

A 4-step agentic workflow pack for coding built to run with ChatGPT, Claude, and Gemini. Open the Markdown files, fill the variables, and paste into your model. Most buyers get a reviewable result in about 15 minutes.

• Catch the runs that finish green and still fail — a 9-entry silent-failure catalog covering empty output, retry-succeeded, stale inputs, swallowed tool errors, and duplicate side effects
• Rank the morning's exceptions on a weighted 5-dimension model where an agent that already acted on the world outranks one that merely stopped
• Get containment separated from repair, so the first fifteen minutes stop the bleeding instead of starting a diagnosis
• Cut the queue against the hours you actually have, with the consequence of every deferred item stated instead of hidden
• Call drift with numeric triggers and a baseline — including the slow slide that never trips a single-day threshold and is how fleets quietly degrade
• Decide escalation from a threshold table, not from how the morning feels, and get the gaps your own policy never covered
• Hand over a shift the way controllers do: changed state, overnight landmines, do-not-touch list, and a carry-forward block that becomes tomorrow's first input
• Pure-prompt playbook — works in any chat model against exports you already have, no integrations required