Skip to main content

Nav Continuity Layer: Outcome Map

Desired Outcome

Starting a Dreamineering build session automatically produces a short briefing (state, current plan, open blockers, last receipts) and loads a small persistent memory feed into the agent's context so work begins in under 30 seconds without re-briefing.

Contributing Factors (with evidence)

  • Extracted Learnings: We already practice extracting learnings into structured artifacts (VVFL/receipts mindset).
  • Ecosystem: There is an existing docs + skills + plans ecosystem to source truth from (.claude/hooks, .agents/skills, docs/).

Obstacles (with mitigations)

  • Multiple sources of truth: (plans, issues, receipts, docs)
    • Mitigation: Define a strict precedence order (Plan → Issues → Receipts → Docs → Git log).
  • Context bloat risk:
    • Mitigation: Enforce hard budgets (e.g., 1-page briefing + last N receipts + last N memory events).

Investigations (No execution until answered)

  • Plan Location: What is the canonical location/format of "current plan" in the repos?
  • Receipts Location: Where do receipts live today, and can we append to nav-memory.jsonl from hooks safely?
  • Commissioning Target: Which agent runtime is the commissioning target first? (Answered: skills and hooks framework natively via KB architecture).

Success Measures (Objective, Binary)

  • On session start, nav-briefing.md is generated automatically.
  • Median "time to productive first message" is under 30 seconds.
  • You no longer paste changelogs/context into chat for 10 consecutive sessions.

Roles (RACI)

  • Accountable: Matt (system owner).
  • Responsible: Nav + engineering agent(s) implementing hook/scripts.
  • Consulted: Docs team (schema + placement), Ops (tooling reliability).
  • Informed: All teams (new startup protocol).

Next Actions (within 48 hours)

  • Decide repo + path for nav-briefing.md + nav-memory.jsonl.
  • Confirm receipt schema file path and a minimal append-only format.
  • Scaffold the prompt deck and engineering spec.

Questions

  • Which startup signal best proves momentum was preserved rather than merely sped up?
  • What is the smallest evidence packet needed to trust the generated briefing?