[prompt-analysis] Copilot PR Prompt Analysis - 2026-10-14

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-10-14

Summary

Analysis Period: Last 30 days
Total PRs: 601 | Merged: 500 (83.2%) | Closed: 88 (14.6%) | Open: 13

Merge rate among completed PRs: 85.0%. No PRs had empty bodies.

Prompt Categories and Success Rates

Categorized by title keywords (first match wins; uncategorized PRs omitted). Success rate = merged / (merged + closed).

Category	Total	Merged	Success Rate
Bug Fix	78	67	89.3%
Feature Addition	95	78	83.0%
Refactoring	14	12	85.7%
Documentation	26	21	87.5%
Testing	10	7	70.0%

Prompt Analysis

✅ Successful Prompt Patterns

• Average prompt length: 164 words (median 107)
• Explicit file references in backticks: 18.4% of merged PRs (92/500)
• Examples: Exclude tool-cache from pre-agent workspace audits #68725 Exclude tool-cache from pre-agent workspace audits; Align MCP server CLI reference with help output #68715 Align MCP server CLI reference with help output; Bump gh-aw-firewall and gh-aw-mcpg dependencies #68598 Bump gh-aw-firewall and gh-aw-mcpg dependencies → all Merged

❌ Unsuccessful Prompt Patterns

• Average prompt length: 200 words (median 100.5)
• Explicit file references: 15.9% of closed PRs (14/88)
• Examples: Scope OTLP credentials to environment-protected telemetry jobs #68555 Scope OTLP credentials to environment-protected telemetry jobs; Add engine max-context-rebuild support and normalize Codex guard hits #68504 Add engine max-context-rebuild support and normalize Codex guard hits; Check Codex safeoutputs availability and remove jq-dependent body submission #68458 Check Codex safeoutputs availability and remove jq-dependent body submission → all Closed

Key Insights

• Length: Closed PRs have a higher mean prompt length (200 vs 164 words) though medians are similar, so the gap is driven by a tail of very long, multi-part prompts.
• Scope: Closed examples bundle several concerns (feature + security + engine behavior); merged ones are narrow, single-purpose changes.
• Category: Bug fixes and docs merge most; testing is lowest, but n=10 is too small to conclude.
• File references show only a weak difference (18.4% vs 15.9%).

Recommendations

1. DO: Keep each prompt to one focused outcome.
2. DO: Frame bug fixes with concrete symptoms and the affected component.
3. AVOID: Long multi-part prompts combining feature work, security changes and engine behavior in one PR.

Historical Trends

Date	PRs	Success Rate	Notes
2026-10-14	601	85.1%	Bug fix highest (89.3%)
2026-10-13	596	85.2%	Docs highest (88.9%)

Trend: Stable; merge rate flat at ~85% and the closed-vs-merged length gap persists.

---

Generated by Copilot PR Prompt Analysis (Run: 38634365874)

Generated by 🔍 Copilot PR Prompt Pattern Analysis · copilot · auto · 15.5 AIC · ⌖ 0.689 AIC · ⊞ 10.9K · ◷

• expires on Oct 15, 2026, 12:48 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 #69105.