[prompt-analysis] Copilot PR Prompt Analysis - 2026-03-16

[github-actions[bot]]

Summary

Analysis Period: Last 30 days | Total PRs: 1,000 | Merged: 803 (80.4%) | Closed: 196 (19.6%) | Open: 1

This is an improvement over the prior week's 78.5% success rate (2026-03-11), continuing an upward trend from 74.4% in early January.

---

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
🐛 Bug Fix	762	622	139	82%
📚 Documentation	19	15	4	79%
🧪 Testing	109	85	24	78%
✨ Feature Addition	71	53	18	75%
🔧 Refactoring	16	12	4	75%
⬆️ Update/Upgrade	10	7	3	70%
❓ Other/Uncategorized	13	9	4	69%

---

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average body length: 931 words (vs 816 for closed — longer prompts correlate with success)
• Average title length: 71 characters
• Most common verbs: fix (353), add (200), update (68), remove (53), refactor (35)
• Conventional commit format: 32% of merged PRs used fix:, feat:, chore: etc. with 84% success rate vs 79% for non-conventional

Example Successful PRs

1. PR #21174: fix: update golden files for gh-aw-firewall v0.24.2 bump — Body describes exact failing test, specific version numbers, and precise change needed → Merged

2. PR #21173: chore: bump gh-aw-firewall default version to v0.24.2 — Precise scope (pkg/constants/constants.go), named constant, exact version strings → Merged

3. PR #21172: refactor: extract shared flattenArtifactTree helper to remove duplicate flatten logic — Named 3 specific functions in a specific file, quantified the duplication (150 lines) → Merged

---

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average body length: 816 words (~12% shorter than merged)
• Vague, unscoped titles with no conventional commit prefix
• Generic bodies like "Pull request created by AI Agent" (no task context)

Example Unsuccessful PRs

1. PR #21143: Completing task — Body: "Pull request created by AI Agent". No task description, no context, no specifics → Closed

2. PR #21113: refactor: extract Daily Audit Discussion safe-output pattern into shared/daily-audit-discussion.md — Well-titled but referenced #[issue number] (placeholder not filled in) → Closed

3. PR #21088: Add /tctt slash command to block users via silent list — Scope may have been too broad or feature not aligned with current roadmap → Closed

---

Key Insights

• Specificity drives success: PRs referencing specific files, constants, or test names in the title/body had 11% file-reference rate in merged vs 7% in closed. The pattern: name the exact thing being changed.
• Conventional commits matter: Using fix:, feat:, chore: etc. yields 84% success vs 79% without — a 5-point advantage. This signals that scoped, intent-clear titles help reviewers quickly assess merit.
• Bug fixes dominate and succeed: 76% of all PRs were bug fixes, with the highest category success rate (82%). The specificity requirements of bug fixes (error message, test name, line number) naturally create better prompts.

---

Recommendations

Based on today's analysis:

1. DO: Use conventional commit format — fix:, feat:, chore:, refactor: in PR titles. They correlate with 5% higher merge rates.
2. DO: Name specific files, functions, constants, or test names in the title and body. "Fix TestWasmGolden_CompileFixtures golden snapshots" is stronger than "fix golden test".
3. DO: Write longer, more detailed bodies (aim for 900+ words). Merged PRs average 931 words; context helps reviewers understand changes.
4. AVOID: Placeholder text like Closes #[issue number] — unfilled placeholders signal incomplete prompts.
5. AVOID: Generic titles like "Completing task" or "Add feature" — these produce the lowest success rates and signal a lack of task clarity.

---

Historical Trends

Date	PRs	Merged	Success Rate	Notable Trend
2026-03-16	1,000	803	80.4%	↑ New high
2026-03-15	1,000	797	80.0%	↑
2026-03-14	1,000	790	79.0%	→
2026-03-13	1,000	784	79.0%	→
2026-03-12	999	787	79.0%	→
2026-03-11	1,000	781	78.5%	↑
2026-03-05	1,000	775	78.0%	baseline

Trend: Steady improvement over the last 2 weeks (+2.4 pts), reaching a new 30-day high of 80.4% today.

---

References: §23128931888

AI generated by Copilot PR Prompt Pattern Analysis · history

• expires on Mar 17, 2026, 5:21 AM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Copilot PR Prompt Pattern Analysis.

A newer discussion is available at Discussion #21340.