Multi-Agent Memory - Build

The build

This build combines the skills below in a fixed order. Required skills are non-negotiable; optional skills add checks without changing the core path.

1
Project Memory
Designs and maintains the shared context files
2
Agent Handoff
Produces state snapshots when sessions end
3
Validation Checklist
Tests whether a fresh session can work from memory alone

Prerequisites

A shared Git repository
Agents configured to load a startup context file
A review process for memory updates

Execution steps

1
Inventory what agents need to know about the project
2
Design the memory files and the startup read order
3
Seed the files with current conventions and decisions
4
Add the decision log and task state tracker
5
Require handoff notes when sessions end mid-task
6
Review memory diffs in the same merge as the code changes
7
Run the fresh-session test after each significant update

Acceptance checklist

A fresh session answers purpose, conventions, and next steps from memory
Decision log entries carry dates, authors, and rationale
No contradictions between memory files
Memory diffs are reviewed with code

Recovery steps

If memory drifts, regenerate it from the current diff

If agents overwrite each other, route updates through review

If files get bloated, audit what agents actually read at startup

What this build includes

Three skills that cover the full memory lifecycle: designing and maintaining context, snapshotting state at handoff, and testing that the whole thing works from a cold start.

When to use it

Use when the cost of agents forgetting the project has become visible: repeated questions, undone work, contradicted decisions. The build converts that hidden cost into a reviewed, diffable file.