Two Hidden Claude Code Commands That Actually Matter

Here's the thing about productivity tools: we download them, use maybe 20% of their features, and convince ourselves we're maximizing efficiency. Claude Code users are no different. Developer and educator Mark Kashef argues there are two slash commands most people either don't know exist or have never bothered to try—and they might be exactly what separates casual users from power users.

The commands in question? /power-up and /insights. Both require Claude Code version 2.1.0.90 or later, and according to Kashef, they serve fundamentally different purposes in the learning curve.

The Learning Companion vs. The Personal Trainer

/power-up is straightforward enough: it's a built-in tutorial system offering 10 interactive lessons covering what Kashef calls "the 80/20 of Claude Code fundamentals." Think of it as onboarding that actually lives inside the tool you're trying to learn.

The lessons walk through tagging files with @-mentions, navigating different operational modes, and working with sub-agents. Kashef demonstrates how each lesson includes ASCII-style visual guides showing exactly what to do. If you've been using Claude Code for a while, this isn't revolutionary. But for newer users, it's the kind of structured learning path that tools often promise in their marketing and rarely deliver in practice.

"This is not going to make you a wizard," Kashef acknowledges. "This will take you from zero to again 80% in terms of core understanding."

That 80% framing is interesting because it sets realistic expectations. Most productivity tools oversell their onboarding features. This one seems designed to get you functional, not expert.

/insights, on the other hand, is where things get more complex—and potentially more useful. Run the command and Claude Code analyzes your last 30 days of usage, generating an HTML report that breaks down how you actually work.

What Your AI Knows About You

The insights report is surprisingly granular. It catalogs your session types, the tools you use most frequently, the programming languages you default to, even your response time patterns. More importantly, it identifies friction points: where you're getting stuck, what features you're ignoring, and which habits are slowing you down.

Kashef walks through his own report, which notes he's built Claude Code into "a genuine business operating system using Gmail, client invoicing, meeting scheduling." The system also flags his mistakes: "wrong diagnosis loops, destructive or over broad actions"—what he calls "basically me lazily prompting."

This kind of self-awareness tool is uncommon in developer software. Most analytics tell you what you did; this one tells you what you could be doing differently. The report includes a section called "On the Horizon" that generates specific workflow recommendations based on your usage patterns.

In Kashef's case, because he tends to use Claude Code early in the morning, the system suggests creating "an autonomous morning operations pipeline" and provides the exact prompt he would need to build it.

The Meta-Automation Loop

Here's where it gets recursive: Kashef demonstrates building a custom skill that runs /insights automatically, extracts the key recommendations, stores them in his Obsidian vault (a popular note-taking app among developer types), and emails him a monthly summary via Google Workspace CLI.

Then he schedules that skill to run automatically using /loop or /schedule.

So you end up with a system that monitors how you use your AI coding assistant, analyzes your patterns, suggests improvements, documents those suggestions in your knowledge management system, and repeats the whole cycle monthly—all without manual intervention.

It's productivity automation eating its own tail. Whether that's genuinely useful or just impressively nerdy depends entirely on how you work and what you're trying to optimize.

The Feedback Loop Question

What's compelling about this setup isn't necessarily the specific tools—Obsidian, Google Workspace CLI, whatever—but the underlying logic. If AI coding assistants are supposed to make developers more productive, shouldn't they also help developers understand their own workflows better?

The /insights command essentially turns Claude Code into a mirror. It shows you your habits, good and bad. The automated pipeline Kashef builds just ensures you actually look at that mirror regularly instead of running the report once, thinking "huh, interesting," and never opening it again.

There's an open question here about whether self-monitoring like this actually changes behavior. Fitness trackers tell you how many steps you took; productivity tools tell you how many GitHub commits you made. Having the data doesn't automatically translate to improvement. You still have to act on the insights.

Kashef's automation addresses this by making the friction points impossible to ignore—they land in your email, they're timestamped in your knowledge base, they're part of your regular workflow review. But that only works if you're the kind of person who actually reads those emails and reviews that knowledge base.

The Accessibility Problem

One thing Kashef glosses over: this entire workflow assumes significant technical literacy. You need to know what a CLI is, how to install Obsidian plugins, how to schedule jobs, how to write prompts that build custom skills.

The /power-up command is accessible to beginners. The /insights automation pipeline? Not so much.

This creates a weird dynamic where the tool most useful for identifying skill gaps requires the most skill to fully utilize. It's the classic problem with meta-tools: the people who need them most are least equipped to use them.

Still, even using /insights manually—just running it once a month and reading the HTML report—could surface blindspots. You don't need the full automation to benefit from seeing your patterns laid out.

Worth the Version Update?

Both commands require updating to Claude Code 2.1.0.90 or later. Whether that's worth it depends on where you are in the learning curve.

If you're new to Claude Code, /power-up is straightforward value: structured lessons, clear progression, built-in guidance. If you're already comfortable with the basics, /insights offers something harder to find—honest feedback about your actual usage, not your imagined ideal workflow.

The automation piece is where it gets more subjective. Building a system that monitors your AI usage and feeds recommendations back into your workflow is either brilliant or overkill, depending on how you think about productivity tools.

What's certain is that these features represent a different approach to AI assistance—one where the tool doesn't just respond to your requests but actively tries to make you better at using it. Whether that's helpful coaching or just another notification to ignore is probably something you'll only know by trying it.

—Zara Chen, Tech & Politics Correspondent