菜单
Check whether llm-tutor has a newer release on GitHub and, if so, surface what's new along with the two commands needed to upgrade. Triggers: '/tutor-update', 'check for llm-tutor updates', 'is there a new tutor version', 'update the tutor', 'pull latest tutor'.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Pick up a paused tutoring session. Finds the user's active in-progress topic, briefly reorients them (where they are, what's acquired, what's coming), and hands the dialogue back to the active persona. If multiple topics are in progress, asks which one. Invoke when you're returning after a break. Triggers: '/tutor-resume', 'continue', 'pick up where I left off', 'back to the tutor', 'resume tutoring'.
Show the llm-tutor dashboard — XP total, cycles balance with time-until-next-refill, active topics with progress, completed topics with XP earned, and lifetime stats. Invoke with no arguments. Triggers: '/tutor-status', 'show me my tutor status', 'how am I doing', 'check my XP', 'where am I at with the tutor'.
Start a Socratic tutoring journey on any user-supplied topic. Generates a learning path of 5-8 prerequisite concepts, calibrates the user's knowledge via 3 graduated prompts, charges 1 cycle, and hands off to the active Echo/Cipher/Vex persona for the dialogue. Invoke with the subject as an argument (e.g. /tutor-start 'python decorators'), or with no argument to be asked. DOES auto-trigger on: 'tutor me on X', 'tutor me through X', 'start a tutoring session on X', '/tutor-start'. DOES NOT auto-trigger on colloquial 'teach me X' or 'explain X' — those usually mean the user wants a quick answer, not a 30-90 minute Socratic dialogue.
Switch llm-tutor statusline icon mode between emoji (default), nerd (Nerd Font glyphs), unicode (geometric text-presentation symbols), and ascii (pure 7-bit fallback). Mirrors claude-statusline's /statusline-icons. Pass an explicit mode name or no argument to cycle. Triggers: '/tutor-statusline-icons', '/tutor-statusline-icons nerd', 'switch tutor statusline to nerd font', 'change tutor statusline icons to ascii'.
Install the llm-tutor statusline integration. Wraps your existing Claude Code statusline (any statusline — claude-statusline, custom, none) so llm-tutor's segment renders as an extra row alongside it. Saves your original statusline command so /tutor-statusline-uninstall can restore it. Triggers: '/tutor-statusline-install', 'install the tutor statusline', 'show my tutor progress in the statusline', 'enable tutor in statusline'.
Toggle the llm-tutor statusline segment on or off without uninstalling the wrapper. When off, the wrapper still runs but only emits the user's original statusline — zero visible change vs. no integration. Triggers: '/tutor-statusline-toggle', '/tutor-statusline-toggle on', '/tutor-statusline-toggle off', 'turn the tutor statusline off', 'hide the tutor statusline', 'show the tutor statusline'.
| name | tutor-update |
| description | Check whether llm-tutor has a newer release on GitHub and, if so, surface what's new along with the two commands needed to upgrade. Triggers: '/tutor-update', 'check for llm-tutor updates', 'is there a new tutor version', 'update the tutor', 'pull latest tutor'. |
This skill answers the question "is there a newer version of llm-tutor than the one I'm running, and what would I get by upgrading?" It compares the installed plugin-cache version against the latest GitHub release, prints the release notes when an upgrade is available, and provides the two exact slash commands the user needs to type to perform the upgrade.
llm-tutor is a Claude Code plugin, so its install/update lifecycle is owned by /plugin update llm-tutor@llm-tutor + /reload-plugins. A skill cannot invoke another slash command, so this skill stops short of performing the update itself. What it does instead is the version check + diff that makes the manual upgrade worth doing — and tells you when you can skip it.
bash "$CLAUDE_PLUGIN_ROOT/scripts/check-for-update.sh"
Echo the script's output verbatim. Don't summarize — the user wants to see the installed version, the latest version, and (when behind) the actual release notes.
If the script reports "You're up to date," no further action is needed. Optionally note the current version.
If the script reports "Update available!", the script already prints the two commands. Add a single line of context:
"Run those two commands above to upgrade — the second one (
/reload-plugins) is required for the new version to take effect in the current Claude Code session without a restart."
If the script reports "(couldn't reach GitHub)", suggest the user check their network and try again, or visit the releases page directly.
/plugin update from this skill. It's a Claude Code built-in slash command and skills can't trigger other slash commands. The user has to type it themselves.~/.claude/plugins/cache/llm-tutor/ directly to "force" an update. That bypasses Claude Code's plugin registry (~/.claude/plugins/installed_plugins.json) and leaves the install in an inconsistent state. The lifecycle is owned by /plugin update.state.json and survives updates — but a reload mid-conversation may discard the active turn's context. Mention this if the user has an active in_progress topic and is mid-dialogue.