Maximize your Claude Code with three skills worth installing, and a prompt for each.

Claude Code. /statusline is built in. The other two skills install from inside a Claude Code session with the /plugin command.

1. 01

   Number three: /statusline. It is built into Claude Code, so there is nothing to install. You describe what you want in plain words, and Claude Code writes a script file in ~/.claude/ and updates your settings. The status line shows in its own row above the built-in footer badges, which stay where they are.

   Status line, in the Claude Code docs

   Send in Claude Code

   Copy prompt

   /statusline identify the 3 highest impact UX changes we can make based on session transcripts and my model correction and steering patterns

   Your part: read what it proposes before you approve a change. This prompt asks for more than a status bar: it has Claude Code look at how you correct and steer the model and propose status line changes from that. In the reel, Claude Code gave it to a background agent that finished in 2 minutes 11 seconds. What comes back depends on your sessions.

2. 02

   Number two: /skill-creator. It comes from the skill-creator plugin in Anthropic's official marketplace. Claude Code adds that marketplace for you the first time you start an interactive terminal session, so one line installs it.

   skill-creator on GitHub, by Anthropic

   Send in Claude Code

   Copy prompt

   /plugin install skill-creator@claude-plugins-official

   Your part: send the line and approve anything Claude Code asks. In VS Code, if the plugin is not found, add anthropics/claude-plugins-official on the Marketplaces tab first.

3. 03

   Run /skill-creator from inside a project. Jason uses it for any workflow that comes down to the same series of actions, repeated over and over. His prompt has it read this project's session transcripts on disk and name the patterns where a skill would save time, improve accuracy or improve outcomes against the project's goal.

   Send in Claude Code

   Copy prompt

   /skill-creator search the session transcripts on disk for this project's history and identify the patterns which emerge that demonstrate a specific skill will save the user time, improve accuracy of work, or enhance outcomes relative to the project's goal

   Your part: open Claude Code in the project you want it to study, then decide which of the skills it suggests are worth building. Transcripts are kept 30 days by default, so it reads about the last month. If the short name clashes with another command, use /skill-creator:skill-creator.

4. 04

   Number one: the CLAUDE.md improver. /claude-md-improver is a skill in Anthropic's claude-md-management plugin, from the same marketplace. It scans for CLAUDE.md files, evaluates them against templates, writes a quality report, then makes targeted updates. It asks for confirmation before it updates anything.

   claude-md-management on GitHub, by Anthropic

   Send in Claude Code

   Copy prompt

   /plugin install claude-md-management@claude-plugins-official

   Your part: send the line and approve anything Claude Code asks. The plugin is named claude-md-management; the skill inside it is claude-md-improver.

5. 05

   Run it in a project first. Project level means ./CLAUDE.md or ./.claude/CLAUDE.md in that repo. The parts in angle brackets are yours to fill in before you send it: the file to audit, who you are, your goals in order, and what done means for your work.

   Fill in the angle brackets, then send in Claude Code

   Copy prompt

   /claude-md-improver Audit <path to CLAUDE.md, or "this repo">.
   
   Context the file should serve:
   - Who I am: <role, company legal name, clients>
   - My goals, in priority order: <1) ... 2) ... 3) ...>
   - What "done" means for my work: <e.g. seen running, rendered, deployed>
   
   How to audit:
   1. Check every path, command, skill, plugin, hook and tool the file names against the
      real system (filesystem, settings.json, enabled plugins, PATH). Flag anything dead.
   2. Find rules that sit at the wrong level: global rules that only fit some repos, and
      repo rules repeated globally.
   3. Check memory files and files the CLAUDE.md links to (commands, sibling CLAUDE.md
      files) for facts that contradict it.
   4. Tell me what the file is missing for my goals above, not only what's wrong with it.
   
   Rules for changes:
   - Back up the file before editing.
   - Show every change as a diff and wait for my OK.
   - You may fix files that directly depend on a change (commands, memories). List each one.
   - Don't commit. Leave third-party and plugin-managed files alone; report their issues.
   - Keep the file as short as or shorter than it is now, unless a line adds what my goals need.

   Your part: replace every <...> with your own details, and leave none in. The prompt tells it to back up the file, show each change as a diff, wait for your OK and not commit. If the short name clashes with another command, use /claude-md-management:claude-md-improver.

6. 06

   Then run it at the global level. That is ~/.claude/CLAUDE.md, the file in your root Claude Code folder (on Windows, %USERPROFILE%\.claude\CLAUDE.md). Claude Code reads it in every project, together with that project's own CLAUDE.md: the files are added to each other, and neither one overrides the other.

   CLAUDE.md levels, in the Claude Code docs

   Fill in the angle brackets, then send in Claude Code

   Copy prompt

   /claude-md-improver Audit ~/.claude/CLAUDE.md.
   
   Context the file should serve:
   - Who I am: <role, company legal name, clients>
   - My goals, in priority order: <1) ... 2) ... 3) ...>
   - What "done" means for my work: <e.g. seen running, rendered, deployed>
   
   How to audit:
   1. Check every path, command, skill, plugin, hook and tool the file names against the
      real system (filesystem, settings.json, enabled plugins, PATH). Flag anything dead.
   2. Find rules that sit at the wrong level: global rules that only fit some repos, and
      repo rules repeated globally.
   3. Check memory files and files the CLAUDE.md links to (commands, sibling CLAUDE.md
      files) for facts that contradict it.
   4. Tell me what the file is missing for my goals above, not only what's wrong with it.
   
   Rules for changes:
   - Back up the file before editing.
   - Show every change as a diff and wait for my OK.
   - You may fix files that directly depend on a change (commands, memories). List each one.
   - Don't commit. Leave third-party and plugin-managed files alone; report their issues.
   - Keep the file as short as or shorter than it is now, unless a line adds what my goals need.

   Your part: fill in the same placeholders. Point 2 of the audit matters most at this level: it looks for global rules that only fit some repos, and repo rules repeated globally.