Beyond Basic Prompts: Dan Sherrard-Smith’s Guide to Mastering Claude

In a recent LinkedIn post, Dan Sherrard-Smith shares a collection of insights and ‘pro tips’ for users looking to get the most out of the AI assistant Claude. Sherrard-Smith emphasizes that many users overlook key functionalities and prompt engineering techniques that can significantly enhance Claude’s performance and output quality. He frames the common issue of generic AI responses not as a failing of the AI, but as a direct reflection of the user’s input.

Sherrard-Smith highlights a common pitfall related to AI honesty, stating:

“Claude lies when you ask ‘say I don’t know if you don’t know.'”

This observation suggests that even advanced AI models may not always adhere strictly to negative constraints when prompted, necessitating more direct or nuanced instructions for accurate responses. He advocates for specificity over vague requests, a theme that runs through several of his points.

Leveraging Claude’s Capabilities for Deeper Understanding

A significant portion of Sherrard-Smith’s advice centers on how to effectively communicate complex information and requirements to Claude. He points out the superiority of detailed system prompts over short, clever phrases, suggesting that Claude benefits from comprehensive context.

Furthermore, Sherrard-Smith recommends utilizing Claude’s file upload feature, noting that the AI genuinely processes the content. He offers a practical tip for verifying this:

“Upload files instead of pasting walls of text. Claude actually reads them. Pro tip: ask Claude how much of the file it read. Sometimes it says it read the file but only got ca. 30% of the way through.”

This underlines the importance of not only providing information but also understanding how the AI interacts with it. When it comes to quality expectations, Sherrard-Smith advises against using subjective terms like “10/10 quality.” Instead, he proposes setting concrete, measurable criteria:

“Asking for ’10/10 quality’ is useless. Ask for specific criteria instead. E.g. ‘No sentences over 15 words. No bullet points. End with a question.'”

This approach, according to Sherrard-Smith, leads to outputs that are far more aligned with the user’s actual needs.

Enhancing Output Through Iterative Refinement and Personalization

Sherrard-Smith also touches on using Claude as a tool for self-improvement and refining one’s own work. He suggests employing the AI to critique ideas before they are finalized or shared publicly.

The Power of Specificity in Prompts

The author stresses that the quality of Claude’s output is intrinsically linked to the quality of the prompt. He states, “If output is generic, your prompt was generic. That’s our issue not Claude’s.” This principle is further illustrated by his preference for constrained explanations:

“Explain it simply in 5 sentences” beats “explain like I’m 5” every time. Short constraints force clarity.

Additionally, Sherrard-Smith champions the use of Claude’s Custom Styles feature, describing it as an underutilized tool for productivity. By uploading samples of one’s writing, users can enable Claude to match their tone and style automatically, thus avoiding repetitive instructions and reducing output-related frustrations.

Other key recommendations include leveraging the mobile app for voice-to-text input, proactively asking Claude for clarifying questions before a task begins, and instructing the AI to challenge assumptions or disagree when necessary. Sherrard-Smith concludes that engaging with Claude in this more sophisticated manner transforms it from a simple answer engine into a powerful collaborative tool.