Nav Continuity Layer: Agent & Instrument Diagram

A&ID
════════════════════════════════════════════════════════════
[TRIGGER: Session Start]
       │
       ▼
   (Agent: Orchestrator Hook)
       │
       ├─► [Instrument: Reader Script] ──► Reads `plan.json`
       │
       ├─► [Instrument: Logger Script] ──► Reads `nav-memory.jsonl`
       │
       ├─► [Instrument: Receipt Parser] ─► Reads `.ai/receipts/`
       │
       ▼
   (Process: Compile Brief)
       │
       ▼
   [Artifact: `nav-briefing.md`]
       │
       ▼
   (Agent: Context Injector)
       │
       ▼
[OUTCOME: Active Agent Session < 30s]
════════════════════════════════════════════════════════════

Nomenclature​

• Agent (Orchestrator Hook): The script or hook executing at IDE/Agent startup (.claude/hooks or .agents/skills framework).
• Instrument (Reader/Logger/Parser): Scripts parsing the specific KB repositories and schemas.
• Protocol (Compile Brief): Translating the raw JSON/YAML state into a human/agent readable Markdown snapshot.

Questions​

Which agent in the nav continuity layer diagram has the highest authority-to-accountability ratio — and is that appropriate?

• Where does human judgment remain essential in this agent-instrument flow, and where is automation safe?
• If one instrument in this diagram reported bad data, which agents would be most affected and how quickly?
• What would need to be true for a new agent type to be added to this diagram without breaking existing flows?