Skip to content

auto-memory.md

Latest commit

 

History

History
53 lines (37 loc) · 1.27 KB

auto-memory.md

File metadata and controls

53 lines (37 loc) · 1.27 KB
Raw
name auto-memory
description End-to-end memory system creation — design, implement, and refine
argument-hint agent-repo-path [user-requirements]
alwaysApply true

auto-memory

Target agent repository: $ARGUMENTS

Goal

Orchestrate the full memory-building pipeline by running each phase in sequence. This is a convenience wrapper — each sub-command can also be run independently.

Pipeline

1. Design

/auto-memory-design <agent-repo> [user-requirements]

Explore the agent and produce memory_design.md.

2. Implement

/auto-memory-implement <agent-repo>

Build the memory system from the design, write the refine adapter, produce implemented.md, and initialize refine_state.json.

3. Refine

/auto-memory-refine <agent-repo>

Run the Critic-Improver loop until the Critic stops or max phases are reached. The command handles the loop internally.

Notes

  • Each sub-command is designed to run in its own session to avoid context drift. If running end-to-end in a single session, be aware that context quality may degrade during later refine phases.
  • Cross-session state is carried by note.md, memory_design.md, implemented.md, and refine_state.json. No information needs to be passed manually between commands.