Added prompt-time recall to memloom connect claude-code: a second hook injects the
memories relevant to each prompt you type, so Claude uses your memory without being told
to search it. Silent when nothing matches or the daemon is down, never blocks a prompt;
opt out with --no-recall (docs)
Added an agent setup guide plus installable setup and usage skills, so an AI agent can
install and configure memloom end to end with the human only supplying the API key and
capture consent (docs)