[prompt-analysis] Copilot PR Prompt Analysis - 2026-10-27

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-10-27

Summary

Analysis Period: Last 30 days (2026-09-28 → 2026-10-27)
Total PRs: 605 | Merged: 507 (83.8%) | Closed: 96 (15.9%) | Open: 2

Completed-PR merge rate: 84.1% (stable vs ~85% in recent runs). All PRs had non-empty bodies.

Prompt Categories and Success Rates

Categorised by keywords in the PR title (a majority of titles use none of these verbs, so they fall in Other).

Category	Total (completed)	Merged	Success Rate
Bug Fix	66	58	88%
Feature Addition	81	69	85%
Refactoring	13	11	85%
Documentation	25	22	88%
Testing	7	5	71% (small sample)
Other / unclassified	411	342	83%

Prompt Analysis

✅ Successful Prompt Patterns

• Average body length: 160 words (merged)
• Bodies that reference concrete files (.go, .cjs, .md, .yml): 81 merged vs 14 closed (85% success)
• Examples: Suppress startup-failure reports for cancelled agent jobs #72405 Suppress startup-failure reports for cancelled agent jobs → Merged; Honor newer cached action pins in runtime setup #72272 Honor newer cached action pins in runtime setup → Merged; Record safe job and script outcomes in unified sessions #72146 Record safe job and script outcomes in unified sessions → Merged

❌ Unsuccessful Prompt Patterns

• Average body length: 138 words (closed)
• Closed PRs skew toward cross-cutting security/infra behaviour changes (bots/permissions, model alias selection, firewall egress) where scope is ambiguous.
• Examples: Allow on.bots Apps to check out same-repository PR edits #72466 Allow on.bots Apps to check out same-repository PR edits → Closed; Exclude Copilot utility models from alias selection and restore detection fallbacks #72404 Exclude Copilot utility models from alias selection → Closed; Separate Agy bootstrap downloads from runtime firewall egress #72102 Separate Agy bootstrap downloads from runtime firewall egress → Closed

Key Insights

• Specific file references help slightly: 85% success with file refs vs 84% without a reference. The difference is small and not statistically strong.
• Longer bodies merge slightly more: merged bodies average 160 words versus 138 for closed, so context appears to help rather than hurt (reverse of an earlier trend where closed prompts were longer).
• Category matters little: all categories sit between 85–88% apart from the tiny testing sample; scope clarity seems to matter more than task type.

Recommendations

1. DO: Include background and the concrete files/behaviour expected, with acceptance criteria.
2. DO: Keep security/firewall/permission changes narrowly scoped, one concern per PR.
3. AVOID: Ambiguous cross-cutting changes touching policy or egress without a defined expected outcome.

Historical Trends

Date	PRs	Success Rate
2026-10-27	605	84.1%
2026-10-25	595	85.0%
2026-10-22	194	85.1%

Trend: Merge rate steady at ~84–85%.

---

Generated by Copilot PR Prompt Analysis (Run: 40579794161)

Generated by 🔍 Copilot PR Prompt Pattern Analysis · copilot · auto · ~14 AIC · ⌖ ~1 AIC · ⊞ 11.6K · ◷

• expires on Oct 28, 2026, 12:45 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 #72753.