Optimize detection thresholds per transaction type, reduce false positives, tune velocity windows, calibrate ML model, improve graph analysis depth

The authoring surface — every primitive this play composes from. Primitives live under .github/ in the source repo and are loaded by the FAI Engine according to the manifest.

• 🤖agents

  3

  Reasoning + tool-calling entities. Each agent declares its model, system prompt, and the skills/tools it composes.

• ⚡skills

  3

  Reusable, multi-step capability definitions. A skill is a folder containing a SKILL.md describing how to compose lower-level primitives.

• 📋instructions

  3

  Behavioural rules scoped by `applyTo` globs. Loaded into every agent that matches the scope.

• 🪝hooks

  1

  Event-driven scripts triggered by lifecycle events (SessionStart, AfterAgentInvoke, OnGuardrailViolation, …).