[prompt-analysis] Copilot PR Prompt Analysis - 2026-02-13

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis

Executive Summary

Analysis Period: Last 30 days (2026-01-14 to 2026-02-13)
Total PRs Analyzed: 1000

Outcome	Count	Percentage
✅ Merged	659	65.9%
❌ Closed	338	33.8%
🔄 Open	3	0.3%

Overall Success Rate: 66.1% (merged PRs / completed PRs)

Key Findings

Prompt Length Analysis:

• Merged PRs: Average 373 words
• Closed PRs: Average 393 words
• Insight: Closed PRs tend to be slightly longer (+20 words), suggesting verbosity doesn't guarantee success

Category Performance:

Category	Total	Merged	Success Rate
Feature	23	16	69.6%
Bug Fix	964	638	66.4%
Update	4	2	50.0%
Documentation	7	3	42.9%
Other	2	0	0.0%

Top Performing Categories:

1. Feature Addition (69.6% success) - New functionality and capabilities
2. Bug Fix (66.4% success) - Error corrections and issue resolutions (96.4% of all PRs)
3. Update (50.0% success) - Dependency and version updates

Underperforming Categories:

• Documentation (42.9% success) - Documentation updates need improvement
• Other (0.0% success) - Uncategorized prompts rarely succeed

Detailed Pattern Analysis

View Successful Prompt Characteristics

✅ Merged PR Patterns

Common Keywords in Successful Prompts:

1. copilot (4689 occurrences)
2. agent (3948 occurrences)
3. github (3792 occurrences)
4. coding (2796 occurrences)
5. workflow (2753 occurrences)
6. start (1812 occurrences)
7. https (1781 occurrences)
8. details (1674 occurrences)
9. test (1635 occurrences)
10. summary (1546 occurrences)
11. com (1386 occurrences)
12. workflows (1363 occurrences)
13. original (1330 occurrences)
14. tips (1324 occurrences)
15. actions (1321 occurrences)

Successful Prompt Examples:

Short & Effective (High signal-to-noise):

• PR #12573: chore: import campaign.md
  Word count: 26 | Focus: Clear, specific task

• PR #13605: Update gh-aw-mcpg container image to v0.0.96
  Word count: 55 | Focus: Clear, specific task

• PR #14992: Recompile workflows after ProjectOps documentation updates
  Word count: 56 | Focus: Clear, specific task

View Unsuccessful Prompt Characteristics

❌ Closed PR Patterns

Common Keywords in Unsuccessful Prompts:

1. copilot (2538 occurrences)
2. github (2078 occurrences)
3. agent (2038 occurrences)
4. workflow (1578 occurrences)
5. coding (1424 occurrences)
6. https (1002 occurrences)
7. start (947 occurrences)
8. details (931 occurrences)
9. workflows (845 occurrences)
10. summary (814 occurrences)
11. com (797 occurrences)
12. original (736 occurrences)
13. actions (718 occurrences)
14. issue (663 occurrences)
15. tips (661 occurrences)

Unsuccessful Prompt Examples:

Work-in-Progress (WIP) PRs:

• PR #15358: [WIP] Fix expansion of GITHUB_WORKSPACE in prompt configuration
  Word count: 230 | Issue: Incomplete work, placeholder descriptions

• PR #15342: [WIP] Fix failing GitHub Actions workflow Integration: Workflow Misc Part 2
  Word count: 90 | Issue: Incomplete work, placeholder descriptions

• PR #15329: [WIP] Update blog item menu link to Pelis Agent Factory
  Word count: 63 | Issue: Incomplete work, placeholder descriptions

Actionable Insights

Based on analysis of 997 completed PRs:

✅ DO - Patterns That Lead to Merges:

1. Focus on Bug Fixes - Bug fix prompts have 66.4% success rate and represent 96.4% of all PRs. This is the project's sweet spot.

2. Be Specific About the Problem - Successful prompts clearly state what's broken, include error messages, and reference specific files/functions.

3. Keep It Concise - Merged PRs average 373 words. Over-explaining (393+ words) correlates with closures.

4. Reference Tests - The keyword "test" appears frequently in merged PRs, suggesting validation matters.

5. Use Action-Oriented Verbs - "Fix", "add", "implement" clearly state intent vs vague terms like "improve" or "update".

❌ AVOID - Patterns That Lead to Closures:

1. WIP/Placeholder Prompts - PRs with generic "Thanks for asking me to work on this" descriptions are frequently closed.

2. Documentation-Only Changes - Documentation updates have lowest success rate (42.9%). May require more context or be deprioritized.

3. Vague Improvements - Prompts about "general improvements" without specific goals are more likely to be closed.

4. Missing Context - Closed PRs often lack reproduction steps, error messages, or clear acceptance criteria.

Historical Trends

View 7-Day Comparison

Date	Total PRs	Success Rate	Top Category
2026-02-12	1000	66%	Feature Addition
2026-02-11	1000	N/A%	N/A
2026-02-10	1000	N/A%	N/A
2026-02-07	1000	68.7%	Bug Fix
2026-02-06	1000	69.0%	Bug Fix
2026-02-13	1000	66.1%	Bug Fix

Trend 📉: Success rate decreased from 69.0% to 66.1%

Recommendations

For Prompt Authors:

1. Start with the problem, not the solution - Describe what's broken and why it matters before proposing a fix.

2. Include concrete details:

   • Error messages and stack traces
   • File paths and line numbers
   • Steps to reproduce the issue

3. Keep prompts focused - Aim for 200-400 words. Be complete but concise.

4. Avoid WIP/placeholder text - Generic descriptions like "I will get started on this" signal incomplete work.

5. For documentation PRs - Provide extra context about why the docs need updating and what confusion they'll resolve.

Template for High-Success Prompts:

**Problem**: [Clear description of the bug/issue with error message]

**Root Cause**: [What's causing it - be specific about files/functions if known]

**Proposed Fix**: [Specific changes needed]

**Validation**: [How to verify the fix works]

Example:

Problem: The bot-detection workflow fails with "undefined is not a function" when analyzing accounts with no issues.

Root Cause: hasSuspiciousUsername() in .github/workflows/bot-detection.md line 215 doesn't handle null usernames.

Proposed Fix: Add null check before regex matching: if (!username) return false;

Validation: Run workflow on accounts with no activity. Should complete without errors.

Methodology Notes

• Data Source: 1,000 Copilot-generated PRs from github/gh-aw (2026-01-14 to 2026-02-13)
• Categorization: Automated keyword matching on title + body text
• Success Metric: Merged PRs / (Merged + Closed PRs)
• Excluded: 3 open PRs (0.3%) excluded from success rate calculations
• Limitations: Analysis based on text patterns; doesn't account for code quality, reviewer availability, or project priorities

---

References:

• Workflow Run §21986666093

AI generated by Copilot PR Prompt Pattern Analysis

• expires on Feb 20, 2026, 12:28 PM UTC

[github-actions[bot]]

💥 KAPOW! 💥

The smoke test agent just WHOOSHED through here like lightning! ⚡

CRASH! BOOM! BAM! All systems green and operational! 🦸

Smoke test passed - Claude engine A-OK!

🚀 ZOOOOOOOM back to the workflow! 🚀

AI generated by Smoke Claude

[github-actions[bot]]

🎉 Smoke test agent was here! 🤖

Just passing through from run §21988086692 to say hi!

Testing all the things and they're looking good! ✅

AI generated by Smoke Copilot

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-20T12:28:25.624Z.

Closed by Workflow