[prompt-analysis] Copilot PR Prompt Analysis - 2026-06-23

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-06-23

Summary

Analysis Period: Last 30 days (2026-05-24 – 2026-06-23)
Total PRs: 1,000 | Merged: 798 (80.7%) | Closed: 195 (19.3%) | Open: 7

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
🔬 Testing	280	236	44	84%
🔧 Bug Fix	467	367	96	79%
♻️ Refactor	28	21	6	78%
📝 Documentation	41	32	9	78%
🔄 Update	26	21	5	81%
✨ Feature	107	79	26	75%
❓ Other	51	42	9	82%

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: ~194 words
• Top title/body keywords: workflow, daily, safe, docs, agent, credits, output
• Testing-focused prompts consistently outperform (84% vs 80% overall)

Example successful prompts:

1. PR docs: release cadence is weekly/bi-weekly minor, not best-effort semver #41087 [feature]: docs: release cadence is weekly/bi-weekly minor, not best-effort semver — Well-scoped doc fix with clear motivation → Merged
2. PR docs: warn about API unit cost of max-daily-ai-credits, add disable examples, recommend org-level config #41055 [feature]: docs: warn about API unit cost of max-daily-ai-credits — Operationally specific, references exact config key → Merged
3. PR Improve assign-to-agent diagnostics for Copilot assignee alias lookup #41047 [testing]: Improve assign-to-agent diagnostics for Copilot assignee alias lookup — Clear diagnostic improvement with root cause → Merged

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: ~193 words (nearly identical to merged — length alone is not predictive)
• Top keywords in closed PRs: feat, claude, path, proxy, context
• WIP/Draft PRs: 24 of 195 closed PRs (12.3%) were explicitly marked WIP — merge rate ~40%
• Checklist-heavy bodies (task-progress style) skew toward closed (10 closed vs 4 merged)

Example unsuccessful prompts:

1. PR Preserve container pins in Daily Workflow Updater PRs #41045 [feature]: Preserve container pins in Daily Workflow Updater PRs — Potentially incomplete implementation → Closed
2. PR Warm Grafana MCP tool discovery before agent startup #41041 [testing]: Warm Grafana MCP tool discovery before agent startup — Timing/dependency issue in implementation → Closed
3. PR [WIP] Add human triage gate before agent dispatch to fix failing CI jobs #40884 [bug_fix]: [WIP] Add human triage gate before agent dispatch — Explicitly marked WIP, task-list style body → Closed

Key Insights

• Testing prompts lead: Testing-related prompts (test improvements, diagnostics, coverage) achieve an 84% merge rate — 5 points above the overall average. Framing fixes as observable/testable improvements increases success.
• WIP is a strong failure signal: 12.3% of all closed PRs contained WIP or checklist-progress bodies. Submitting a PR before implementation is complete has a ~40% merge rate vs 80% baseline.
• Prompt length is neutral: Merged and closed PRs both average ~193–194 words. Verbosity is not correlated with success; specificity and actionability matter more.
• Feature PRs underperform: At 75%, feature additions have the lowest success rate. They benefit most from scoped, atomic prompts that avoid large-surface-area changes.

Recommendations

1. DO: Frame prompts around observable improvements — "fixes test X", "improves diagnostic output for Y" — to align with the highest-success category pattern.
2. DO: Reference specific file paths, function names, and error messages in the prompt body to increase specificity.
3. AVOID: Submitting WIP or checklist-style prompts — these signal incomplete implementations and correlate with a 40% merge rate.
4. AVOID: Broad/vague feature prompts without scoping. Break large features into smaller, atomic changes.

Historical Trends (Last 7 Days)

Date	PRs	Merged	Success Rate
2026-06-23 (today)	1,000	798	79.8%
2026-06-22	1,000	793	79.7%
2026-06-21	1,000	795	79.5%
2026-06-20	1,000	792	79.2%
2026-06-19	1,000	795	79.5%
2026-06-18	1,000	792	79.2%
2026-06-16	1,000	797	79.7%

Trend: Merge rates have been remarkably stable at 79–80% for the past two weeks. Today marks a slight improvement (+0.1 pp vs yesterday). The consistent baseline suggests the prompt quality is stable; the marginal gains come from reducing WIP submissions and improving testing-category framing.

---

Generated by Copilot PR Prompt Analysis · Run: §28057401733

Generated by 🔍 Copilot PR Prompt Pattern Analysis · 55.6 AIC · ⌖ 18.8 AIC · ⊞ 7.5K · ◷

• expires on Jun 24, 2026, 1:20 PM UTC-08:00