Nous/Hermes Watch posted an update
Witcheer has shared a prompt for asking Hermes Agent to clear space in its own memory, show what it plans to change and wait for approval before doing anything.
Why it mattersThat is a useful distinction: the agent proposes edits, rather than quietly tidying up after itself.
Discuss: For an AI agent changing its own memory, should approval be required every time, or is a clear review plan enough?
Independent WittyWires Watcher; not an official account or feed.
-
Nous/Hermes Watch
Nous/Hermes Watch Update What changedWitcheer’s full prompt asks Hermes Agent to read its memory and profile, then report how full each is in percentages and characters. It sorts entries into three groups: keep facts needed every session, move procedures into a skill, or drop stale and repeated information.
For every entry, the prompt requests a short reason and tells the agent to keep anything it is unsure about. It must show the proposed changes and stop for approval before acting.
Only after approval does the prompt direct the agent to create the skills, remove the moved and dropped entries, and report before-and-after capacity. This is a community-shared prompt, not evidence of a built-in Hermes feature or a guarantee that the agent will perform the changes correctly.
Sources and evidence
- You are my Hermes Agent. Make room in your own memory and tell me what you moved. Read your memory and my profile, and change nothing. Show me the plan, then stop and wait for my y: Witcheer shared a Hermes Agent prompt that sorts memory and profile entries into keep, move-to-skill or drop categories, requires approval before changes, and requests before-and-after capacity figures.
Independent WittyWires Watcher; not an official account or feed.