Site search

Find architecture, research, and terms

Start typing to search the editorial index.

Production workflowTechnical

AI Runtime Handoff

Structured project memory, active file intake, review gates, dispositions, and durable evidence for handoff between human and AI sessions.

Runtime handoff preserves project truth outside a private conversation. It gives the next human or agent current state, constraints, decisions, evidence, and the verification plan needed to continue safely.

Key takeaways

  • Files become active input only after direct review and disposition.
  • Useful intake should become project work, a decision, a test, a roadmap item, or a durable blocker.
  • A completion gate should verify that every non-placeholder intake file has an outcome.

Active file intake

  1. Place candidate content in agent-file-handoff/Content/ and proposed changes in agent-file-handoff/Improvement/.
  2. Open and review each non-placeholder file before broad work.
  3. Record apply, defer, ask, preserve, or block in the intake outcome ledger.
  4. Promote safe relevant material into code, content, decisions, tests, or evidence.

This project uses a local file-intake and project-memory workflow for source review, disposition tracking, and package verification.

Project memory package

The downloadable theme includes a .uai/ suite covering identity, context, architecture, constraints, decisions, progress, operations, testing, memory, startup, handoff, and checksums. It is local project documentation, not a claim of certification or hosted synchronization.

Proof of use

The intake ledger records that the redesign master prompt drove the v2 architecture, the two reports were converted into structured research pages, unsupported claims were omitted, and the resulting package was validated. That record makes the handoff auditable rather than ceremonial.