[prompt-clustering] 🔬 Copilot Agent Prompt Clustering Analysis - November 18, 2025 #4249
Closed
Replies: 1 comment
-
|
This discussion was automatically closed because it was created by an agentic workflow more than 1 week ago. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
🔬 Copilot Agent Prompt Clustering Analysis - 2025-11-18
This analysis uses Natural Language Processing (NLP) and K-means clustering to identify patterns in Copilot agent task prompts from the last 30 days. By analyzing 984 PRs, we discovered 7 distinct task clusters with varying success rates and characteristics.
Summary
Analysis Period: Last 30 days
Total Tasks Analyzed: 984
Clusters Identified: 7
Overall Success Rate: 77.2%
Key Insights
The analysis reveals that bug-fix tasks, particularly those focused on code, workflow, and package modifications (Cluster 2), achieve the highest success rate at 83%. Feature tasks dominate in volume with 439 total tasks across multiple clusters, while update tasks show strong performance at 79% success rate.
Full Analysis Report
Cluster Analysis
Cluster 5: Feature Tasks
Cluster 7: Feature Tasks
Cluster 2: Bug-Fix Tasks
Cluster 6: Bug-Fix Tasks
Cluster 4: Bug-Fix Tasks
Cluster 3: Update Tasks
Cluster 1: Bug-Fix Tasks
Success Rate by Cluster
Sample Data (First 50 PRs)
Recommendations
Based on clustering analysis:
High Success Pattern: Bug-Fix tasks (Cluster 2) show 83% success rate. Keywords: pkg, workflow, code. Consider this as a model for well-defined tasks.
Task Diversity: Identified 7 distinct task patterns. This diversity suggests the agent handles a wide range of work types effectively.
Feature Task Patterns: Feature tasks split into two main clusters - general additions (Cluster 5) and workflow-specific features (Cluster 7). Both show similar success rates around 76%.
Update Tasks: With 79% success rate, update tasks perform well, suggesting that modification requests with clear scope are effective.
Analysis Date: 2025-11-18
Dataset: 984 PRs from last 30 days
Method: TF-IDF vectorization + K-means clustering (k=7)
Silhouette Score: 0.075 (indicates moderate cluster separation)
Beta Was this translation helpful? Give feedback.
All reactions