mubit-ai
OrganizationOfficial Mubit marketplace: persistent, typed, self-improving memory for Claude Code and the Codex CLI.
Categories
Indexed Skills (15)
recall
Search Mubit memory for lessons, rules, facts or past work; use when the injected memory is not enough.
activity
List or export what this instance holds for a run; use when the user wants the audit record, not an answer.
auth
Sign in to Mubit and store an API key; use when the plugin reports auth_failed or a key was rotated.
checkpoint
Save a named verbatim snapshot of this run; use when the user is about to do something risky.
dashboard
Open the local Mubit dashboard — turns, what was injected into each, what it earned, and every lesson's record; use when the user wants to look, not ask.
doctor
Diagnose connectivity, memory health and stuck ingest; use when memory looks empty or the status line shows a failure.
forget
Delete a lesson, or down-weight one that is merely wrong; use when asked, after saying it cannot be undone.
handoff
Hand work to another agent through Mubit, list the handoffs open in this run, or answer one with a verdict. Use when a task changes hands between agents, or when a subagent's result is waiting for review.
import
Backfill Mubit memory from the Codex rollouts and Claude Code transcripts already on this machine, so a fresh install knows what happened before it. Use when the user asks to import, backfill or seed their history, or when memory is empty because the plugin was installed after the work. Sends nothing without an explicit --send.
memory-health
Report what is actually stored: counts, staleness, contradictions; use when the question is whether anything was ever stored.
pin
Pin a constraint for the rest of this run, shown on every prompt; use when a rule ends with the task, and remember for anything durable.
reflect
Extract lessons from this session's activity now; use when the user wants them banked before session end or asks what memory holds.
remember
Save one durable lesson, rule or preference; use when the user states a standing preference or an approach fails in a way worth never repeating.
setup
First-run setup: check the endpoint, the key and that the instance answers; use when freshly installed or reporting auth_failed.
strategies
Surface the pattern across many stored lessons; use when the user asks what memory has learned in general or why a mistake keeps recurring.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.