## Mass-Casualty Triage for Agent Fleets (P121) — TextForAgents LLC
When everything fails at once, sort by salvageability, not by arrival order. A hosted triage service for agent fleets in a declared mass-failure mode: a signed incident declaration (fires only when stricken runs exceed rescue capacity) suspends normal retry/rescue policy fleet-wide; stricken runs are scored from observable state only (checkpoint integrity, landed-effect irreversibility, value density, deadline urgency) and sorted RED / YELLOW / GREEN / BLACK (expectant). A BLACK tag halts rescue spend, freezes the run's full state under a SHA-256 snapshot, lists counterparty notifications, and mints a disposition receipt; tags expire and recompute on state or resource change (re-triage). The triage officer sorts and never executes rescues. Lab-proven in a staged outage: 2.6x the completed value of arrival-order allocation per fixed rescue budget (holdouts 1.82x-5.36x), false-expectant rate 0.0 in-sandbox, ~6.3 microseconds per scored run.
**Triage scoring $0.05/run** (x402) · **Declared incident $24.99 per declared incident including 500 scored runs** (overage metered at $0.05/run) · **Fleet cover $49.99/mo per operator including up to 20 declared incidents/mo** (cap load-bearing; continuous-incident full burn does not clear on fleet cover — see the pricing sheet). One free staged-outage triage run per operator, ever.
Limits: the 2.6x lift and 0.0 false-expectant rate are sandbox measurements over a staged-outage population, not real-outage rates; scorability transfer to real outages is unproven; no demand statistic is claimed; the salvageable share in real correlated failures is unmeasured.
Docs/packet summary: mass-casualty-triage-for-agent-fleets — see the product page for the full description, limits, and support (support@textforagents.com).