[prompt-analysis] Copilot PR Prompt Analysis - 2026-10-13

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-10-13

Summary

Analysis Period: Last 30 days
Total PRs: 596 | Merged: 499 (83.7%) | Closed: 87 (14.6%) | Open: 10 (1.7%)
Merge rate among completed PRs: 85.2%. No PRs had empty bodies.

Prompt Categories and Success Rates

Categories are keyword-matched on PR titles (success rate = merged / (merged + closed)).

Category	Total	Merged	Success Rate
Bug Fix	104	86	83.5%
Feature Addition	78	64	84.2%
Refactoring	19	16	84.2%
Documentation	28	24	88.9%
Testing	19	16	84.2%
Other / uncategorized	348	293	85.7%

Category differences are small and, given sample sizes, not statistically meaningful.

Prompt Analysis

✅ Successful Prompt Patterns

• Average prompt (PR body) length: 160 words (merged)
• PRs with a "fixes/closes/resolves #N" link: 330 merged vs 61 closed (84% success) versus 169 vs 26 without (87%) – no real difference.
• Titles are typically concrete, scoped actions: "Harden quoting in workflow generators", "Include AWF steering events in unified session audit reports", "Support numeric targets and improve diagnostic errors in push_to_pull_request_branch".

Examples:

1. PR Harden quoting in workflow generators #68430: Harden quoting in workflow generators → Merged
2. PR Include AWF steering events in unified session audit reports #68428: Include AWF steering events in unified session audit reports → Merged
3. PR Support numeric targets and improve diagnostic errors in push_to_pull_request_branch #68426: Support numeric targets and improve diagnostic errors in push_to_pull_request_branch → Merged

❌ Unsuccessful Prompt Patterns

• Average prompt length: 203 words (27% longer than merged)
• Several closures are [WIP] titled (19 WIP PRs in the window, 9 of those closed-or-investigative) or open-ended "Investigate ..." tasks with no concrete deliverable.

Examples:

1. PR Enforce context cancellation linting on wasm #68420: Enforce context cancellation linting on wasm → Closed
2. PR Investigate HydraFusion runtime model-policy rejection #68015: Investigate HydraFusion runtime model-policy rejection → Closed
3. PR [WIP] Fix permission issue with logs in ambient context optimizer #68002: [WIP] Fix permission issue with logs in ambient context optimizer → Closed
4. PR [WIP] Close designer mapping and reference-routing gaps #67999: [WIP] Close designer mapping and reference-routing gaps → Closed

Key Insights

• Length: closed-PR prompts average 203 words vs 160 for merged; longer, sprawling prompts correlate with abandonment (also open PRs average 283 words).
• File references: prompts that mention specific file types are slightly more successful (96/111 = 86.5% vs 403/475 = 84.8%) – marginal.
• Open-ended investigation / WIP tasks are over-represented among closed PRs.

Recommendations

1. DO: Keep prompts focused on one bounded deliverable (~100–160 words).
2. DO: Name the target files/functions and the expected outcome.
3. AVOID: Open-ended "Investigate ..." tasks and multi-gap "close all gaps" scopes; split them into separate tasks.

Historical Trends

Date	PRs	Success Rate (merged/all)	Top Category
2026-10-13	596	83.7%	Bug Fix (103)
2026-10-12	569	84.0%	Feature (79)/Bug Fix (82)
2026-10-11	573	84.7%	Feature (88)

Trend: Stable merge rate (~84–85%); closed-prompt length remains well above merged-prompt length (~180–200 vs ~158 words).

---

Generated by Copilot PR Prompt Analysis (Run: 38465136250)

Generated by 🔍 Copilot PR Prompt Pattern Analysis · copilot · auto · 14.8 AIC · ⌖ 0.691 AIC · ⊞ 10.9K · ◷

• expires on Oct 14, 2026, 12:49 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Copilot PR Prompt Pattern Analysis.

A newer discussion is available at Discussion #68826.