@suluk/agents - v0.1.6
    Preparing search index...

    Interface AgentGradeOptions

    interface AgentGradeOptions {
        instructions?: Record<string, string>;
        catalog?: ModelCatalog;
        modelWindows?: Record<string, number>;
        served?: string[];
        snapshots?: Record<string, string>;
    }
    Index

    Properties

    instructions?: Record<string, string>

    instruction snapshots keyed "<agent>/<skill>" — lets the context analyzer MEASURE instruction load (else lower-bound).

    catalog?: ModelCatalog

    the @suluk/models catalog — enables the model-fit dimension (window vs estimated peak load).

    modelWindows?: Record<string, number>

    per-id context-window overrides (tests/pins; takes precedence over the catalog).

    served?: string[]

    the tools a server actually advertises by default — folds in the over-serve + cold-tail-in-default conformance checks.

    snapshots?: Record<string, string>

    the CURRENT served instruction snapshot, keyed qualified "<agent>/<skill>" (wins) OR bare "<skill>" (back-compat) — same dual-accept as instructions + verifyAgentFreshness; folds in the skill-freshness (drift) check.