[prompt-analysis] Copilot PR Prompt Analysis - 2026-04-11

[github-actions[bot]]

Summary

Analysis Period: Last 30 days (2026-03-19 → 2026-04-11)
Total PRs: 1,000 | Merged: 813 (81.5%) | Closed: 184 (18.4%) | Open: 3

⚠️ Trend Alert: Success rate declined to 81.5% this week, down from 83.2% last week — the lowest in the tracked period.

---

Prompt Categories and Success Rates

Category	Total	Merged	Success Rate
🔼 Upgrade / Bump	2	2	100%
🚀 Feature Addition	90	79	88.8%
🧪 Testing	12	10	83.3%
🐛 Bug Fix	876	707	80.9%
📄 Documentation	14	11	78.6%
🔧 Refactor	3	2	66.7%

Bug fixes dominate (87.6% of all PRs) — but feature additions achieve the highest success rate among substantial categories.

---

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average body length: ~850 words
• Top title prefixes: fix: (326 merged), feat: (112 merged), add (54 merged), docs: (42 merged)
• Conventional commit format strongly associated with merged PRs
• 86% of fix: prefixed PRs were merged
• 85% of feat: prefixed PRs were merged

Example merged PRs:

1. PR #25750: fix: normalize container pins in wasm golden tests → Merged ✅
2. PR #25746: Normalize report formatting instructions across daily workflow files → Merged ✅
3. PR #25745: Fix test failures after action pin updates → Merged ✅

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average body length: ~896 words (slightly longer — correlates with over-scoped tasks)
• WIP PRs: 56 total, 49 closed (87.5% failure rate) — largest single predictor of failure
• [WIP] prefix appeared in 47 closed vs only 6 merged
• Closed PRs often involve broader architectural changes, alternative approaches, or speculative features

Example closed PRs:

1. PR #25696: Replace resolve_host_repo.cjs with job.workflow_repository / job.workflow_ref... → Closed ❌
2. PR #25686: feat: add --no-ask-user flag to copilot agentic engine for fully autonomous runs → Closed ❌
3. PR #25660: refactor: split Go BLOCKER files exceeding 1,000 lines to resolve architecture v... → Closed ❌

View WIP PR Analysis

Out of 56 [WIP] tagged PRs:

• Merged: 6 (10.7%)
• Closed: 49 (87.5%)
• Open: 1

This is the single strongest predictor of a closed PR. WIP PRs were closed at a rate nearly 4.5x higher than the overall closed rate. Avoid submitting PRs tagged as WIP unless they are explicitly intended for early feedback.

Top closed WIP examples:

• [WIP] Fix Design Decision Gate 100% failure due to empty prompt
• Multiple speculative refactoring attempts

---

Key Insights

• 🏆 Insight 1: Feature addition prompts outperform bug fix prompts (88.8% vs 80.9%). Prompts that introduce clear new capabilities tend to succeed more often than those targeting unclear or diffuse bugs.
• ⚠️ Insight 2: [WIP] is a near-certain failure signal — 87.5% of WIP PRs were closed. If the work is exploratory, either complete it before opening a PR or mark it as a draft.
• 📉 Insight 3: Success rate has declined week-over-week (88% → 81% → 81% → 78% most recent). Week 15 (Apr 7–11) shows 78% — higher than average closure volume (48 closed PRs this week). This may reflect increased speculative or experimental PR volume.

---

Recommendations

Based on today's analysis:

1. DO use conventional commit prefixes (fix:, feat:, add, docs:) — they correlate with 85–89% success rates and signal clear intent
2. DO keep prompts focused on a single, well-scoped change — feature addition PRs (88.8% success) tend to be more targeted than bug fix catch-alls
3. AVOID marking PRs as [WIP] — 87.5% of WIP PRs were closed; complete the work or use draft PRs instead
4. AVOID overly broad refactoring requests — architectural refactors in the sample had the highest closure correlation

---

Historical Trends

Date	PRs	Merged	Success Rate	Notable
2026-04-01	1000	832	83.5%	Baseline
2026-04-02	1000	823	82.5%
2026-04-03	1000	824	82.7%
2026-04-05	1000	825	82.8%
2026-04-06	1000	828	83.0%
2026-04-07	1000	828	83.2%	Peak
2026-04-11	1000	813	81.5%	⚠️ Lowest

Trend: Success rate has declined 1.7 percentage points from the April 7 peak. The increase in closed PRs (+17 vs last week) warrants monitoring — may reflect a batch of exploratory or experimental work.

---

References: §24274980340

Generated by Copilot PR Prompt Pattern Analysis · ● 1.3M · ◷

• expires on Apr 12, 2026, 4:54 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-04-12T04:54:34.578Z.

Closed by Workflow