[prompt-analysis] Copilot PR Prompt Analysis - 2026-03-17

[github-actions[bot]]

Summary

Analysis Period: Last 30 days | Dataset: 1,000 PRs from copilot/* branches
Total PRs: 1,000 | Merged: 812 (81.2%) | Closed: 185 (18.5%) | Open: 3 (0.3%)

Overall success rate is 81.2%, up slightly from 80.3% yesterday.

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
🐛 Bug Fix	759	629	127	82.9%
✨ Feature Addition	175	129	46	73.7%
♻️ Refactoring	47	38	9	80.9%
🧪 Tests	5	5	0	100%
📚 Docs	1	0	1	0%
Other	13	11	2	84.6%

Conventional Commit Verb Success Rates

Verb	Merged / Total	Rate
chore	14/14	100%
ci	6/6	100%
perf	4/4	100%
refactor	30/34	88%
docs	37/44	84%
fix	219/256	85%
feat	45/60	75%

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average body word count: 921 words (median: 580)
• Use of conventional commit prefixes: fix:, refactor:, chore:, ci:, perf:
• Scoped commits with context: e.g., fix(audit): return absolute paths in downloaded_files
• Specific file/function references in title and body

Example successful PRs:

1. PR refactor: extract shared MCP renderer helpers across engine implementations #21336: refactor: extract shared MCP renderer helpers across engine implementations → Merged
2. PR fix(ci-coach): fallback to issue when PR touches protected files #21333: fix(ci-coach): fallback to issue when PR touches protected files → Merged
3. PR test(fileutil): expand coverage to all exported functions #21332: test(fileutil): expand coverage to all exported functions → Merged
4. PR fix(audit): return absolute paths in downloaded_files #21331: fix(audit): return absolute paths in downloaded_files → Merged

View Closed PR Examples (Patterns to Avoid)

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average body word count: 826 words (median: 430) — slightly lower than merged
• Vague or high-level titles (e.g., Remove PAT-dependent automatic lockdown, replace with default guard policies)
• Missing conventional commit prefix (e.g., Completing task with only 6 words in body)
• Large architectural changes that may be too broad

Example closed PRs:

1. PR Completing task #21143: Completing task — body only 6 words → Closed (insufficient description)
2. PR feat: pin microsoft/apm CLI version with user override and auto-update monitoring #21288: feat: pin microsoft/apm CLI version with user override and auto-update monitoring → Closed
3. PR Remove PAT-dependent automatic lockdown, replace with default guard policies #21279: Remove PAT-dependent automatic lockdown, replace with default guard policies → Closed
4. PR Replace GitHub MCP server lockdown with gateway guard policy #21273: Replace GitHub MCP server lockdown with gateway guard policy → Closed

Key Insights

• Scoped conventional commits win: Prompts yielding fix(scope): or feat(scope): PRs have higher success rates than bare imperatives without scope. Scoping signals specificity.
• Bug fixes dominate and succeed: 75.9% of all PRs are bug fixes, with 82.9% success — the most reliable category. This suggests Copilot performs best when given a clear, concrete problem to fix.
• feat: underperforms vs fix:: Feature additions succeed at only 73.7% vs 85% for bug fixes. New features likely require broader context, stakeholder alignment, or design decisions that exceed prompt scope.
• Minimal body = risk signal: PRs with very short bodies (< 50 chars) trend closed (3 closed, 5 merged). A well-described body signals intent and increases review acceptance.

Recommendations

1. DO: Use conventional commit format with scope: fix(component): specific problem — structured titles strongly correlate with merge success.
2. DO: For bug fixes, include reproduction details, the affected component, and expected behavior in the prompt — bug fix PRs have the highest volume and solid 83% success rate.
3. DO: For refactoring/chore/CI tasks — these have ≥88% success rates. Prefer incremental, well-scoped changes over large rewrites.
4. AVOID: Vague imperative titles without conventional commit prefixes (e.g., Remove X, replace with Y without refactor: prefix).
5. AVOID: Prompts that drive large architectural changes in a single PR — these correlate with closure.

Historical Trends (Last 7 Days)

Date	Total PRs	Merged	Success Rate
2026-03-17	1,000	812	81.2% ↑
2026-03-16	1,000	803	80.3%
2026-03-15	1,000	797	79.7%
2026-03-14	1,000	790	79.0%
2026-03-13	1,000	784	78.4%
2026-03-12	999	787	78.8%
2026-03-11	1,000	781	78.1%

Trend: 📈 Steady improvement — success rate has climbed from 78.1% to 81.2% over the past 7 days (+3.1 pp). Daily PR volume is consistent (~30–47 new PRs/day active days).

Daily Activity Breakdown (Last 10 Days)

Date	New PRs	Merged	Closed	Daily Rate
2026-03-17	10	8	0	80% (day in progress)
2026-03-16	37	31	6	84%
2026-03-15	42	35	6	83%
2026-03-14	30	25	5	83%
2026-03-13	47	38	9	81%
2026-03-12	21	18	3	86%
2026-03-11	37	31	6	84%
2026-03-10	25	22	3	88%
2026-03-09	20	15	5	75%
2026-03-08	24	21	3	88%

References:

• §23179274103

AI generated by Copilot PR Prompt Pattern Analysis · history

• expires on Mar 18, 2026, 5:11 AM UTC