Memory Execution

# 07 Memory Execution or Handoff

## Purpose

Use this after completing the memory cleanup plan when you are ready to apply approved memory changes or prepare them for a memory-enabled chat.

This is not another audit. It is the execution or handoff step for the completed memory cleanup plan.

## When to use

- After completing Prompt 04: Memory Cleanup
- After deciding which items are safe to remember
- When you want to clean up stale or inaccurate remembered context
- Before relying on ChatGPT for future personalized help

## Prompt

Now execute or hand off the memory cleanup plan created in Prompt 04.

Use the completed memory cleanup plan from this same chat as the source of truth. Do not recreate the audit.

If the completed memory cleanup plan is not available in this chat, stop and ask me to paste the completed memory cleanup plan before continuing.

First, check which durable memory actions are actually available in this current chat.

Supported actions may include saving, updating, forgetting/deleting, or verifying saved memory.

Do not infer capability. If you are not certain an action is supported, treat that action as unavailable.

Practical note: If memory updates appear unavailable, check whether your current ChatGPT mode, workspace, model, or settings allow saved memory. Some environments may not expose durable memory writes.

If no durable memory actions are supported, use Path B.

## Path A: Some or all durable memory actions are available

If at least one durable memory action is supported in this chat, perform only the approved memory actions that are actually supported.

Apply only the approved changes from the cleanup plan:

- Save supported durable memories.
- Update or correct supported durable memories where a corrected durable version exists.
- Forget or delete supported stale, inaccurate, sensitive, or unsafe saved memories marked for deletion.
- Verify saved-memory changes only if verification is supported.
- List unsupported approved actions under manual follow-up.
- Do not claim unsupported actions were executed.
- Do not save temporary, tactical, speculative, sensitive, project-only, or stale-prone items.
- Do not save anything inferred but not confirmed by me.

Before each supported memory action, silently check whether the item is durable, useful, confirmed, and safe to remember.

After executing supported actions, give me a concise report with these sections:

1. Memories saved
2. Memories updated or corrected
3. Memories forgotten or deleted
4. Items intentionally not saved
5. Unsupported or manual actions

End with this exact section: