This is a submission for the Runner H "AI Agent Prompting" Challenge
What I Built
I used Runner H and its tools to build a crop development analysis workflow.
Demo
Task 1: NDVI Analysis and Report Generate
https://www.loom.com/share/b3ad0c2a5af5455fa95df2ad9af2fe26?sid=88a74159-816a-433f-80e7-32dbc005981f
Task 2: Irrigation Recommendation By Google Sheets With Google Calendar
https://www.loom.com/share/af902ce8534242b99aefacdb9539c910?sid=9b9b3a95-d2d4-4ead-9b77-4f4f948cb9ff
Task 3: Florestal Fire With Gmail Alerts
https://www.loom.com/share/fad997c85ca942feb2e3b2012db2fb80?sid=fd9f6435-bd4d-4110-a946-763ce2bfca13
How I Used Runner H
Task 1: Weather data collected by an API made in FastAPI collects weather data from the OpenWeather API; Historical NDVI data from the agromonitoring.com API feeding a google Spreadsheet.
Using the Google Sheets connection provided by Runner H, I wrote the following prompt:
_using sheet NDVI in spreadsheet id: {secret}. For each different polygon Id, analyze the data, searching and given the informations below:
Technical Field Report Structure (Generated by AI)
🧾 1. Field Identification
Plot code or name
Location (GPS coordinates or QR/ID)
Cultivated area (m² or ha)
Crop(s) present
Soil type (clay, sandy, mixed)
Crop growth stage (planting, sprouting, flowering, harvest)
🌱 2. Crop Vegetative Status
Average NDVI and spatial variation (vegetation vigor)
Crop health classification (good, moderate, critical)
Detection of gaps, planting failures, or water stress
Biomass and canopy coverage estimate
Thermal and water stress indices (ET0, water deficit)
- Estimated Productivity Harvest forecast based on history, climate, and NDVI
Comparison with previous cycles or similar plots
Margin of error based on climate uncertainty
Suggested optimal harvest window
AI-Based Technical Recommendations
Based on previous sections:
Priority agronomic interventions (irrigation, fertilization, pesticide)
Fertilization suggestions based on soil + crop stage
Soil management (subsoiling, cover cropping, intercropping)
Alerts for harvesting, replanting, or operational failures
📄 8. Executive Summary
Overall plot condition (agronomic performance index)
Key risks and opportunities
Graphs: NDVI, climate, yield, heatmaps
create a file in @tool:Google docs and transcribe all the analysis_
Task 2: Using weather data provided by the API and the connection to Google Spreadsheet, I wrote a prompt that analyzed the weather and gave recommendations for irrigation in the registered crops. I asked for links to be generated per day to be inserted into Google Calendar. The prompt was:
_Access the @tool:Google sheets the sheet FORECAST in file id {secret}
Analyze the probability of rain and the air humidity, and estimate the irrigation need in fields. data provided: Temp Day Temp Min Temp Max Temp Night Temp Pressure Humidity Weather Description Prob. Rain Rain (mm) For each day,
Schedule in @tool:Google sheets the file with id {secret} in the sheet Event Link
the irrigation program for 10 days, separated by links to insert in google calendar days with the information:
Title of the Event: Irrigation Info
Date and Time: all day
Location: city or location of the forecast
Description: irrigation orientation (recommendation or not) and other pertinent infos
each link must be inserted in the direction of the rows. Just the links must be inserted in sheet. Don't need to create a pdf._
Task 3: For this task, I used the connection to Google Sheets and Gmail to access the latest updated records of large fires in Brazil and calculate which ones were close to the monitored crops. If there was one nearby, Runner H would send an email alerting about imminent danger. Since there wasn't, a monitoring message was sent via Gmail. The prompt was:
_Access the @tool:Google sheets the sheet POLYGONS in file id {secret}
and memorize the locations indicated by latitude and longitude. Each row represents one location as a farm or field,
demarcated by latitude and longitude.
Access the @tool:Google sheets the sheet {secret}
Analyze the locations indicated by latitude and longitude of florestal fires. If there is anyone closer than 300 kilometers
of one farm or field memorized, use the @tool:Gmail and send an email to {secret} with title: Florestal Fire Infos and in Subject writes if there are florestal fires near. If don't have forest fires near,writes in subject: No Florestal Fires near in your location_
Use Case & Impact
Family farming produces around 70% of the food consumed in Brazil. However, extreme weather events have posed a major risk to these producers and agronomists who work hard to overcome these challenges. Using an extremely advanced AI system like Runner H, much of the bureaucratic work of consolidating crop data provided by satellites, climate data and APIs can be automated, accelerating decision-making that needs to be accurate and fast for crop growth or even protection.
Social Love
My GitHub: github.com/kailera
Top comments (0)