Ruben Hassid Shares 9 Claude Formulas for Enhanced AI Workflows

In a recent LinkedIn post, Ruben Hassid shares a comprehensive set of nine formulas designed to optimize the use of Claude, an AI assistant, for professional workflows. Hassid emphasizes that these are not suggestions but mandatory guidelines for his team, aiming to streamline interactions and improve the quality of AI-generated output.

Hassid introduces the concept of transforming static text files into dynamic, callable skills within Claude. He explains the inefficiency of repeatedly uploading a general ‘about-me’ text file for every task. Instead, he proposes a method to convert such information into a custom skill.

“Turn this into a skill I invoke with /about-me. Trim it under 2,000 words. Ask me questions”

This approach, according to Hassid, ensures that Claude re-reads and utilizes the information only when specifically invoked, saving processing time and improving focus for subsequent tasks. He provides a downloadable file via his guide to assist users in setting this up.

Optimizing Claude Interactions with Skills and Projects

A core principle Hassid advocates for is the strategic use of “Skills” and “Projects.” He defines a Skill as something teachable to another person, and a Project as a client or campaign. The real power, he argues, comes from combining these elements.

Hassid illustrates this with an example: using a “contract” skill within a specific client’s project, integrated with Gmail. This allows Claude to draft a contract in the user’s style, incorporating the client’s actual context from the email thread.

“You can teach it to someone → Skill. It’s a client or a campaign → Project. The move nobody does: combine them.”

This integration, as Ruben Hassid points out, moves beyond generic outputs to highly personalized and context-aware document generation.

Streamlining Claude’s Settings and Prompting Techniques

Further refining Claude’s utility, Hassid advises a minimalist approach to settings. He recommends turning off “Memory” and leaving “Global Instructions” empty, while keeping “Connectors” active.

Hassid argues that excessive context can lead Claude to produce repetitive or circular answers. This simplified setting is ideal for brainstorming and creative tasks where flexibility is key, contrasting with situations requiring consistent output, like weekly contract drafting.

Encouraging Proactive Questioning from AI

A significant portion of Hassid’s advice focuses on making Claude a more active participant in the task. He suggests a prompt structure designed to make the AI ask clarifying questions before proceeding.

“I want [task] to [success criteria]. Ask me questions using AskUserQuestion before you start.”

According to Ruben Hassid, this technique is highly effective in preventing AI hallucinations and ensuring the task aligns perfectly with user intent. He notes that if Claude falters mid-conversation, instructing it to “ask more questions” can resolve the issue by forcing it to seek necessary clarification.

Cost-Effective Model Usage and Efficient Editing

Hassid also addresses the practicalities of model selection and cost management. He differentiates between Claude’s Fable-5 and Opus models, recommending Fable-5 for lower-effort tasks and switching to the more capable Opus 4.8 for complex stages.

He explains that Fable-5 operates on usage credits, which can become expensive for extended sessions. By strategically using Opus, which is often included in subscription plans, users can manage costs effectively while leveraging the most powerful features when needed.

Ruben Hassid strongly advises against iterative follow-ups and suggests using Claude’s “Edit” function instead. He highlights that each new message incurs costs and requires Claude to re-process previous context.

“So never type ‘no, I meant…’ Click Edit on your prompt. Everything after it disappears.”

This direct editing approach, as Hassid details, is far more efficient than adding corrective messages, saving both time and computational resources.

Additional Efficiency Formulas

Hassid rounds out his advice with three bonus formulas aimed at further enhancing efficiency:

• Batching Tasks: Combining multiple related prompts into a single message to reduce Claude’s re-reading overhead.
• Voice Input: Utilizing voice-to-text tools like wispr.ai to dictate prompts, leveraging a faster speaking rate (150 words/minute) compared to typing (60 words/minute).
• Task Categorization: Using “Chat” for quick questions, “Code” for software development, and “Cowork” for document-based tasks.

These nine formulas, as presented by Ruben Hassid, offer a practical toolkit for anyone looking to maximize their productivity and the effectiveness of their interactions with AI assistants like Claude.