5 Lesser‑Known but Highly Useful Claude Code Slash Commands

Claude Code’s slash commands are the most frequently used feature for daily interaction. Common commands include /model to switch models, /usage to view quota, /context to inspect conversation history, /mcp for MCP configuration, and /skills to list installed skills. As Claude Code evolves, new slash commands and many Skills become available, expanding the command set.

1. /insights – a health check of your Claude Code usage

Running /insights without arguments generates an HTML report that analyses recent local sessions. The report lists the projects you worked on, problems encountered, and suggests under‑used features. Unlike /usage, which shows raw quota consumption, /insights evaluates how you use Claude Code and offers improvement tips.

Key limits: a single run can analyse up to 200 new sessions; very short sessions are skipped. The report is saved to ~/.claude/usage-data/report.html and a timestamped copy can be kept. It works only for local session diagnostics.

The author ran /insights on a month’s worth of sessions and received a detailed health‑check report covering strengths, issues, immediate fixes, and longer‑term efficiency suggestions.

2. /subtask – dispatch a sub‑agent and receive only the result

/subtask <task description>

The official description: “Send a subagent off with your full context; its result comes back here.” Unlike a regular sub‑agent that starts with a blank slate, /subtask inherits the full conversation history, so the task description can be brief.

Two main benefits: (1) full context inheritance, eliminating the need to repeat background information; (2) only the final result is returned, while the sub‑agent’s intermediate files stay in its own context.

The author cites an Anthropic blog on session cost, which recommends keeping exploratory tasks separate from the main conversation. /subtask implements this best practice. Example: while evaluating a MICCAI competition in the main session, the author used /subtask to let Claude check the competition website and forum, returning the findings without polluting the primary context.

Comparison with /fork: /fork clones the current session into a new background session that runs independently, whereas /subtask sends the result back to the original session.

3. /artifact-design – create and deliver design artifacts

/artifact-design

is a skill‑type slash command. It produces an “artifact” – a web page on claude.ai that contains the design output, has a unique URL, is private by default, and remains permanently accessible. Artifacts do not interfere with the ongoing conversation and can be managed later via the /artifacts command.

The author previously wrote a detailed guide (linked) on Claude Code artifacts. The command is especially handy for research, solution design, or post‑project summaries, delivering results in a smooth, “silky” manner.

It also includes a preset design checklist: beige background (#F4F1EA), serif headings, terracotta accents, near‑black background with fluorescent‑green highlights, purple‑blue gradient hero image, Inter or Space Grotesk as the body font, emojis for section markers, centered layout, rounded‑corner cards, and automatic disabling of these AI‑style defaults when the user does not request a style.

4. /dataviz – data‑driven visualisation

/dataviz

triggers a design guide for charts and dashboards and automatically calls /artifact-design to deliver the visualisation as an artifact.

The official definition lists supported media: HTML charts, inline SVG, Matplotlib, Plotly, D3 code, or PNG images.

The author tested the command with the Kaggle “Kobe Bryant career shooting” dataset, producing a polished visualisation report that can be filtered by season and even link to specific game footage. Screenshots of the generated visualisations are included.

5. /design – invoke Claude Design directly from Claude Code

/design

connects Claude Code to the early preview of Claude Design (claude.ai/design). It allows users to generate professional design drafts without leaving the Claude Code environment.

The author links to a separate article discussing the limited window for AI‑design vendors after Claude Design’s release.

In summary, the author recommends using /subtask frequently, while /dataviz, /artifact-design, and /design serve design‑related needs; /dataviz focuses on visualisation, /artifact-design on front‑end artifact delivery, and /design on full‑fledged professional design. Running /insights monthly provides a usage health check to continuously improve workflow.