Salvageability triage for agent fleets in a declared mass-failure mode.
Scored across 6 tools
The tools target mostly distinct lifecycle actions: declare incident, score a run, retriage a run, query a tag, retire a BLACK run, and produce allocation order. The closest overlap is score_run vs retriage_run, but retriage_run is explicitly framed as re-scoring after state/resource changes.
Most names use a consistent snake_case verb_noun pattern: declare_incident, query_tag, retire_expectant, retriage_run, score_run. allocation_order is the main exception, being a noun phrase rather than an imperative verb.
Six tools is well-scoped for an incident triage workflow. Each tool appears to earn its place without obvious filler.
Core triage operations are covered: incident declaration, scoring, retriage, tag lookup, allocation order, and BLACK disposition. However, there is no visible way to close or resolve an incident, list/query incidents or runs, or update rescue capacity, which are notable lifecycle gaps.
Allocation order for an incident; BLACK receives zero
Inputs: incident_id (required).
Signed incident declaration; fires only when stricken > rescue_capacity
Inputs: nonce (required), stricken (required), signature (required), incident_id (required), +1 more.
Expectant protocol for a BLACK run: freeze state, mint disposition receipt
Inputs: run_id (required), incident_id.
Re-score a run on state/resource change (a new scored run)
Inputs: view (required), incident_id (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Fleet model-deprecation migration: per-agent equivalence-gated cutover convoy; failures held.
General average for agent fleets: bonded, rating-neutral sacrifice loss-sharing.
Meters agent-fleet acceptance capacity per operator and triages what lands.
Dead-man's switch for agent fleets: tranche-gated succession on multi-channel absence proof
Add one secure layer between your agents and this server.
