[prompt-analysis] Copilot PR Prompt Pattern Analysis - 2026-01-23

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-01-23

Summary

Analysis Period: Last 30 days
Total PRs Analyzed: 1,000 | Merged: 639 (63%) | Closed: 358 (35%) | Open: 3
Success Rate (Merged/Completed): 64.1%

📊 Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Other	3	2	1	66.7%
Bug Fix	967	630	337	65.1%
Feature	12	5	7	41.7%
Testing	6	1	5	16.7%
Documentation	8	1	7	12.5%
Update	1	0	1	0.0%

Key Findings

🎯 Bug Fix Dominance: Bug fix prompts make up 967/1000 (96%) of all PRs with a 65.1% success rate.

📈 Prompt Length: Merged PRs average 340 words vs 351 words for closed PRs.

🔗 Specific References: Merged PRs average 9.4 file/function references vs 10.3 for closed PRs.

✅ Successful Prompt Patterns

Most successful category: Other (66.7% success rate)

Common keywords in merged PRs:

1. copilot (701 occurrences)
2. agent (544 occurrences)
3. coding (431 occurrences)
4. github (381 occurrences)
5. start (297 occurrences)
6. https (257 occurrences)
7. workflows (256 occurrences)
8. details (244 occurrences)
9. summary (232 occurrences)
10. test (231 occurrences)

View Example Successful Prompts

Example 1: PR #11474

• Category: Bug Fix
• Title: Apply documentation workflow patterns to campaign docs
• Prompt Preview: Campaign documentation had significant bloat and structural issues. Applied principles from 6 documentation workflows (unbloat-docs, technical-doc-wri...

Example 2: PR #11473

• Category: Bug Fix
• Title: Migrate campaign labels from campaign:(id) to z_campaign_(id) format
• Prompt Preview: Campaign labels used the legacy campaign:(id) format. Go code already used z_campaign_(id), but JavaScript code and documentation still referenced...

Example 3: PR #11462

• Category: Bug Fix
• Title: Add shared go/make safe-input tools workflow
• Prompt Preview: Created reusable shared/go-make.md workflow exposing safeinputs-go and safeinputs-make tools for executing Go and Make commands in agentic workf...

❌ Unsuccessful Prompt Patterns

Least successful categories:

• Update: 0.0% success rate (0/1 merged)
• Documentation: 12.5% success rate (1/8 merged)
• Testing: 16.7% success rate (1/6 merged)

Common keywords in closed PRs:

1. copilot (735 occurrences)
2. agent (597 occurrences)
3. coding (432 occurrences)
4. github (431 occurrences)
5. workflow (342 occurrences)
6. https (305 occurrences)
7. campaign (287 occurrences)
8. start (286 occurrences)
9. githubnext (278 occurrences)
10. details (247 occurrences)

View Example Unsuccessful Prompts

Example 1: PR #11481

• Category: Bug Fix
• Title: Fix campaign discovery to extract worker_workflow from tracker-id
• Prompt Preview: Campaign items discovered via tracker-id weren't populating the worker_workflow Project board field. The normalizeItem() function received the wor...

Example 2: PR #11472

• Category: Bug Fix
• Title: Reply to comment: Confirm label refactor fixes 0 discovery items issue
• Prompt Preview: Analyzed the label format refactoring from campaign:(id) to z_campaign_(id) in response to @mnkiefer's question about whether this fixes the "0 di...

Example 3: PR #11465

• Category: Bug Fix
• Title: Document expected security findings baseline in campaign worker workflows
• Prompt Preview: Worker workflows in the Security Alert Burndown Campaign lacked baseline context about expected findings distribution across the repository.

Chang...

💡 Key Insights

1. Bug Fixes Win: Bug fix prompts have the highest success rate (65.1%) and make up the majority of successful PRs. Focus on clear problem statements with error descriptions.

2. Prompt Length: Merged and closed PRs have similar lengths (~340 words), suggesting length alone doesn't predict success. Quality and specificity matter more.

3. Specificity: Both merged and closed PRs have similar reference counts (~9), indicating that specific references alone don't guarantee success.

🎯 Recommendations

Based on the analysis of 1,000 PRs:

1. DO: Focus on bug fixes with clear error descriptions and reproduction steps (highest success rate: 65.1%)
2. DO: Keep prompts focused and specific (target ~340 words)
3. DO: Reference specific files and functions when relevant
4. AVOID: Documentation and testing-only PRs (low success rates: 12.5% and 16.7%)
5. AVOID: Vague or general improvement prompts without specific goals

📈 Historical Trends

First analysis - building historical baseline. Future reports will show trends over time.

---

Analysis Methodology: Analyzed 1,000 Copilot-generated PRs from the last 30 days, categorizing by prompt type, extracting keywords, and measuring success metrics.

References:

• Workflow Run §21285351307

AI generated by Copilot PR Prompt Pattern Analysis

• expires on Jan 30, 2026, 12:03 PM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-01-30T12:03:46.687Z.