← ClaudeAtlas

importlisted

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.
mubit-ai/plugins · ★ 3 · AI & Automation · score 76
Install: claude install-skill mubit-ai/plugins
**This skill uploads work history to the configured Mubit instance.** It runs one Node process from the plugin directory, reads `~/.codex/sessions` — and, when asked, Claude Code's `~/.claude/projects` — and ingests what it finds. Nothing is sent unless `--send` is on the command line. **Do not run it on your own initiative, and do not pass `--send` unless the user said to.** The Claude Code copy of this skill enforces that with `disable-model-invocation: true`; Codex has no such key, so here it is a rule rather than a mechanism. It is a stricter rule than the one on `activity` and `dashboard`, which only read. This one takes months of somebody's transcripts — every prompt, every command, every file they touched — and puts them on a server. ## Step 0 — resolve the binary No environment variable carries this plugin's path under Codex. Codex lists each skill with the absolute path of its `SKILL.md`; **this file** is at `<plugin-root>/skills/import/SKILL.md`, so the binary is two directories above it: ``` <plugin-root>/bin/import.mjs ``` Resolve that to an absolute path from this file's own location and use it in every command below. Do not write `${CLAUDE_PLUGIN_ROOT}`: Codex sets no plugin-root variable of any spelling, so the shell expands it to nothing and `node /bin/import.mjs` fails with ENOENT. ## Always dry-run first, and show them the number ```bash node <plugin-root>/bin/import.mjs ``` That is the whole command, and it sends nothing: a dry run is what happens w