[prompt-analysis] Copilot PR Prompt Pattern Analysis - 2026-02-04

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-02-04

Summary

Analysis Period: Last 30 days (1000 PRs analyzed)
Total PRs: 1000 | Merged: 679 (68.0%) | Closed: 320 (32.0%)

Key Finding: [WIP] Titles Strongly Correlate with Closure

🚨 Critical Pattern Discovered: PRs with [WIP] in the title are 11.7x more likely to be closed than merged:

• Closed PRs: 31.6% have [WIP] in title
• Merged PRs: Only 2.7% have [WIP] in title

This suggests that incomplete or experimental PRs rarely reach completion.

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Remove/Delete	168	128	40	76.2% ⭐
Refactoring	178	129	49	72.5% ⭐
Updates	560	385	175	68.8%
Testing	593	406	187	68.5%
Documentation	553	376	177	68.0%
Feature Addition	871	591	280	67.9%
Bug Fix	790	530	260	67.1%

Top performers: Remove/delete operations (76.2%) and refactoring tasks (72.5%) have the highest success rates.

Prompt Characteristic Analysis

What Makes Prompts Successful?

Characteristic	Merged PRs	Closed PRs	Advantage
Code References (files, functions)	99.7%	90.9%	+8.8 pp
Issue References (#12345)	39.8%	51.9%	-12.1 pp ⚠️
Reproduction Steps	40.4%	45.3%	-4.9 pp
Error Messages	15.2%	16.2%	-1.0 pp
[WIP] in Title	2.7%	31.6%	-28.9 pp 🚨

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: 416.8 words (median: 316 words)
• Code specificity: 99.7% reference specific code (files, functions)
• Completion status: 97.3% do NOT have [WIP] in title
• Top keywords: copilot, agent, github, workflow, coding, test
• Action verbs: fix, add, update, remove, implement

View Example Successful Prompts

1. PR #13684: Fix cache race condition in parallel integration tests

   Parallel integration test jobs were failing during cache restoration with "File exists" tar errors. All jobs shared the same cache key, causing extrac...

2. PR #13677: Fix gosec G104 violations in MCP config renderers

   Security linter failing on unhandled error returns from RenderJSONMCPConfig() in MCP configuration renderers.

Changes

• Explicitly discard erro...

3. PR #13668: Remove validation, use "latest" as default version for install script

   The install script fails in environments where api.github.com is blocked by MITM proxies (e.g., GitHub Copilot agent runtime, corporate networks), r...

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: 420.5 words (median: 345.0 words)
• Work-in-progress: 31.6% have [WIP] in title (vs 2.7% for merged)
• Over-referencing issues: 51.9% reference issue numbers (vs 39.8% for merged)
• Top keywords: copilot, agent, github, workflow, coding, actions
• Completion rate: Only 68% of non-WIP PRs get merged

View Example Unsuccessful Prompts

1. PR #13686: Add "Use Shared Agentic Workflows" subsection to create.md

   The documentation lacked discovery paths for shared workflow components like orchestration and monitoring patterns. Users had no clear entry point to ...

2. PR #13681: [WIP] Fix failing GitHub Actions workflow for CLI completion

   Thanks for asking me to work on this. I will get started on it and keep this PR's description up to date as I form a plan and make progress.

(!-- ST...

3. PR #13680: [WIP] Investigate toc for design patterns sections
   ) ## ToC Grouping for Design Pattern Sections

Investigating whether Monitoring and Orchestration design pattern sections should be in a separate ToC gr...

Key Insights

Based on analysis of 1,000 Copilot PRs from the last 30 days:

1. 🎯 Avoid [WIP] Titles: PRs marked as work-in-progress are 11.7x more likely to be closed. Start with a clear, complete prompt rather than iterative development.

2. 📝 Simple > Complex: Remove/delete operations (76.2%) and refactoring (72.5%) succeed more than bug fixes (67.1%) or features (67.9%). Simpler, well-scoped tasks perform better.

3. 💻 Code Specificity Matters: 99.7% of merged PRs reference specific files or functions, compared to 90.9% of closed PRs. Being explicit about what to change improves success.

4. 🔗 Issue References Don't Help: Surprisingly, closed PRs reference issue numbers more often (51.9%) than merged PRs (39.8%). Linking to issues doesn't guarantee success.

5. 📏 Length Doesn't Matter: Merged and closed PRs have similar average lengths (417 vs 421 words). Verbosity isn't the deciding factor—clarity is.

Recommendations

Based on pattern analysis, follow these best practices:

✅ DO:

1. Be specific about files and code: Reference exact file names, function names, and line numbers
2. Use clear, completion-ready titles: Avoid [WIP], [Draft], or [Experimental] prefixes
3. Scope tasks narrowly: Smaller, well-defined changes (like removals or refactoring) succeed more often
4. Focus on the what and why: Explain what needs to change and why, with specific examples

❌ AVOID:

1. [WIP] or work-in-progress titles: 31.6% of closed PRs have WIP markers—start with clear intent
2. Over-reliance on issue references: Linking to issues doesn't replace clear problem description
3. Vague prompts without code references: 90.9% of closed PRs lack specific file/function references
4. Large, multi-faceted changes: Complex features and bug fixes have lower success rates

📋 Recommended Prompt Template

[Clear, specific title without WIP]

## Problem
[Describe the specific issue in 1-2 sentences]

## Proposed Solution
[Explain what needs to change, referencing specific files/functions]

## Files to Modify
- `path/to/file.go`: [What to change and why]
- `path/to/test.go`: [Related test changes]

## Expected Outcome
[Clear success criteria]

Historical Trends

This is the first analysis run. Historical trends will appear in future reports as data accumulates.

---

Analysis of 1,000 Copilot-generated PRs from the last 30 days
Generated: 2026-02-04
Run: §21670919805

AI generated by Copilot PR Prompt Pattern Analysis

• expires on Feb 11, 2026, 12:16 PM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-11T12:16:49.492Z.

Closed by Workflow