---
title: "Generic Agent Prompts"
description: "Universal prompts for any MCP-compatible agent."
---
If you are building your own agent or using a client not listed here, these prompts follow a "tool-first" structure that works across most LLMs.
## The "Search Discoverability" Check
Ensures your content is being found.
```markdown
1. List all sitemaps for https://example.com.
2. Check for any sitemap errors.
3. Find any pages in the sitemap that are NOT currently indexed by using the URL inspection tool on a sample of 5 pages.
```
## The "Portfolio Overview"
Useful if you manage multiple properties.
```markdown
1. List all sites in my Search Console account.
2. For each site, get the total clicks and impressions for the last 7 days.
3. Which site had the highest percentage growth in non-brand traffic?
```
## The "Health Check" Prompt
Monitor the health of one or all sites.
```markdown
1. Run a health check across all my Search Console properties.
2. For each site, report the status (healthy / warning / critical) and any issues found.
3. For sites with critical status, use the drop attribution tool to check for algorithm update correlation.
4. Provide a ranked list of sites that need attention, with specific next steps for each.
```
## The "Cross-Engine Comparison"
Compare Google and Bing performance for any site.
```markdown
1. Use the compare_engines tool to compare Google and Bing data for https://example.com over the last 28 days.
2. List the top 10 keywords with the biggest ranking gap between the two engines.
3. Flag any keywords with a 'bing_opportunity' or 'google_dependency_risk' signal.
4. Summarize which engine is underperforming and suggest 3 actions.
```
## Common Tips for Custom Agents
* **Specify Dimension:** When asking for analytics, specify if you want it by `query`, `page`, `country`, or `date`.
* **Specify Engine:** Most tools accept an `engine` parameter (`google` or `bing`). Use `compare_engines` to query both at once.
* **Limit your scope:** LLMs work best when they process chunks of data. Instead of "Analyze everything," use "Analyze my top 50 pages."
* **Ask for reasoning:** Always include "Explain why you reached these conclusions" at the end of your prompt to see the agent's logic.
