Custom Instructions

Custom Instructions let you define persistent preferences that shape how ModelMonkey interacts with you. Instead of repeating the same context in every conversation, you set it once and it applies automatically to all future sessions.

Think of it as giving ModelMonkey background knowledge about you - your role, your industry terminology, how you prefer data formatted, and how detailed you want explanations.

To configure your custom instructions:

1. Open the Menu - Click the three-dot menu icon in the top right corner of the ModelMonkey sidebar.

2. Select Custom Instructions - Click "Custom Instructions" from the dropdown menu.

3. Enter Your Instructions - Write your preferences in the text area. Be specific about what matters to you.

4. Save - Click the Save button to apply your changes.

Your instructions take effect immediately for new conversations. Existing sessions continue with their original context.

Effective custom instructions typically cover one or more of these areas:

Your Role and Expertise
Tell ModelMonkey who you are and what you work on. This helps it calibrate explanations and assumptions appropriately.

Formatting Preferences
Specify how you want data displayed - currency formats, decimal places, date formats, percentage notation, or number separators.

Communication Style
Indicate whether you prefer concise bullet points or detailed explanations, technical language or plain English, formal or conversational tone.

Domain Terminology
Define industry-specific terms, company acronyms, or specialized vocabulary that ModelMonkey should understand and use correctly.

Workflow Preferences
Describe how you typically work with spreadsheets - the types of analyses you run, the reports you create, or the data structures you use.

I'm a financial analyst working with LBO models and valuations.
Use GAAP standards, format currency with thousand separators
($1,234.56), percentages with 1 decimal (12.5%), and dates
as MM/DD/YYYY. Keep explanations concise and data-driven.

Here are practical examples for different use cases:

Marketing Manager
"I analyze marketing data but I'm not highly technical. Use simple language, start with key insights, include actionable recommendations, and highlight unusual patterns in data."

Data Scientist
"I'm comfortable with statistics and Python. You can use technical terms freely. When suggesting formulas, explain the mathematical rationale. I prefer efficiency over verbosity."

Operations Manager
"I track inventory and supply chain metrics. Use metric units, format dates as DD-MMM-YYYY, and always include variance analysis when comparing periods. Flag anything that deviates more than 10% from historical averages."

Sales Team Lead
"I manage a sales team and track quota attainment. Format all revenue as USD with no decimals. Use fiscal quarters (Q1-Q4 starting February). Highlight top and bottom performers in any ranking."

Custom instructions have the following constraints:

Character Limit
Instructions are limited to 4,000 characters. This is enough for comprehensive preferences while keeping context manageable.

Minimum Length
Instructions must be at least 10 characters to be valid. Brief instructions like "be concise" work fine.

Content Validation
Instructions are validated for safety. Content that attempts to override system guidelines, extract sensitive information, or manipulate ModelMonkey's core behavior will be rejected. Stick to genuine preferences and you won't encounter issues.

Editing Instructions
Return to the Custom Instructions page anytime to update your preferences. Changes apply to new conversations immediately.

Clearing Instructions
Click the Clear button to remove your custom instructions entirely. Confirm the action in the dialog that appears. This resets ModelMonkey to default behavior.

Canceling Changes
If you've made edits but haven't saved, click Cancel to revert to your previously saved instructions.

Be Specific
"Format numbers nicely" is vague. "Use thousand separators and 2 decimal places for all currency values" is clear and actionable.

Prioritize What Matters
Focus on preferences that actually affect your work. You don't need to cover every possible scenario.

Test and Iterate
After saving instructions, try a few prompts to see how ModelMonkey responds. Refine your instructions based on what works.

Keep It Relevant
Stick to spreadsheet-related preferences. Instructions about unrelated topics won't improve your experience.

Update as Needed
Your preferences may change as you work on different projects. Revisit your instructions periodically to keep them current.

Custom instructions work best for stable, long-term preferences. For one-off context, include it directly in your message instead.

Use Custom Instructions For:

• Your role and expertise level
• Standard formatting preferences
• Communication style
• Common terminology

Use Per-Message Context For:

• Specific project details
• One-time data requirements
• Temporary formatting needs
• Context unique to a single task

The two approaches complement each other. Custom instructions provide your baseline preferences, while per-message context handles situation-specific needs.