Lawyer Uses Claude Skills, Legal World Loses It

Key Takeaways

• Lawyer built custom Claude “skills” encoding personal workflow.
• Claude edits .docx XML directly, preserving formatting.
• Solution outperforms generic legal AI and reduces manual tasks.
• Demonstrates AI’s ability to replace specialized legal tech tools.
• Highlights shift toward individual AI customization in law firms.

Summary

Lawyer Zack Shapiro showcased how he leverages Anthropic’s Claude “Skills” to automate contract review and formatting tasks at his boutique firm. By creating custom instruction files that embed his decade‑long analytical framework, Claude can edit Word documents at the XML level, apply personalized citation standards, and generate track changes without external plugins. The approach delivered expert‑level output in a fraction of the time, outperforming mainstream legal‑tech solutions. The post has amassed over 7 million views, sparking industry debate about AI‑driven individualization versus firm‑wide playbooks.

Pulse Analysis

The legal sector has long wrestled with fragmented software solutions that address isolated tasks—citation formatting, document numbering, or version control—without integrating into a lawyer’s unique workflow. Anthropic’s Claude, a large language model, now offers a "Skills" framework that lets practitioners encode their own analytical templates, voice, and procedural preferences directly into the AI. By converting these instructions into executable code, Claude becomes a personalized assistant rather than a generic tool, bridging the gap between broad AI capabilities and the nuanced demands of legal drafting.

From a technical standpoint, Claude’s ability to manipulate .docx files at the XML layer is a game‑changer. Instead of relying on third‑party macros or manual copy‑pasting, the model writes precise markup that Microsoft Word recognizes, preserving tracked changes, styles, and cross‑references automatically. This eliminates the time‑draining formatting errors that traditionally consume hours of billable work. Early adopters report turnaround times cut by up to 80%, allowing lawyers to focus on substantive analysis rather than repetitive software chores, and delivering outputs that are indistinguishable from senior associate labor.

The broader implication is a shift toward AI‑driven individualization within law firms. As more practitioners develop proprietary "skills," the competitive landscape may pivot from firm‑wide technology stacks to personal AI expertise. This could accelerate the adoption of generative AI across transactional practice, pressure traditional legal‑tech vendors to offer deeper customization, and raise ethical considerations around AI‑encoded judgment. Ultimately, Shapiro’s experiment illustrates how AI can rewrite the economics of legal services, delivering higher efficiency while redefining the skill set required of modern attorneys.