[prompt-analysis] Copilot PR Prompt Analysis - 2026-06-27

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-06-27

Summary

Analysis Period: Jun 6–27, 2026 (last 30 days)
Total PRs: 1,000 | Merged: 809 (81%) | Closed: 187 (19%) | Open: 4

📈 Trend: Merge rate improved from 79.6% (Jun 10) → 81.2% today — highest recorded in this run series.

---

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Documentation	24	20	3	87% 🥇
Testing	109	92	17	84%
Bug Fix	833	672	159	81%
Feature	15	12	3	80%
Refactor	16	11	4	73%
Other	3	2	1	67%

Note: Bug Fix dominates the dataset (83% of all PRs) — this repo is heavily focused on correctness-driven iteration.

---

Action Verb Success Rates

Action Verb	PRs	Merge Rate
bump/update	29	90% 🥇
fix	236	86%
doc/readme	36	83%
improve/strengthen	24	83%
refactor	33	82%
add/implement/create	151	79%
gate/remove/retire	17	76%
test (standalone)	3	67%

---

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average body length: ~194 words / 1,563 characters
• 84% include code blocks (vs 69% in closed PRs — a 15pp gap)
• Concrete action verbs: fix, bump, restrict, extract, gate, strengthen
• Specific titles with component scope: e.g. fix(changeset): restrict safe-outputs..., refactor: extract shared CLI engine...

Example successful PRs:

1. PR fix(changeset): restrict safe-outputs allowed-files to .changeset/*.md #41937: fix(changeset): restrict safe-outputs allowed-files to .changeset/*.md → Merged — scoped fix with precise file pattern
2. PR refactor: extract shared CLI engine execution logic; make Pi backend resolution configurable #41921: refactor: extract shared CLI engine execution logic; make Pi backend resolution configurable → Merged — specific refactor with clear deliverable
3. PR Strengthen anomaly detector edge coverage and remove order-dependent test behavior #41916: Strengthen anomaly detector edge coverage and remove order-dependent test behavior → Merged — test improvement with concrete problem statement

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average body length: ~183 words / 1,489 characters (slightly shorter)
• 65% of [WIP]-tagged PRs are closed (26/40 — vs 37% for non-WIP)
• Vague or exploratory titles, missing concrete deliverables
• Keywords correlating with closure: wip, proxy, path, phase 1, placeholder

Example unsuccessful PRs:

1. PR [WIP] Restore fleet-wide token_usage collection after 6 days of gap #41814: [WIP] Restore fleet-wide token_usage collection after 6 days of gap → Closed — placeholder body, not ready
2. PR [WIP] Add Sentrux PR gate to CI for quality floor protection #41808: [WIP] Add Sentrux PR gate to CI for quality floor protection → Closed — early WIP with incomplete work
3. PR [q] Shift Typist execution plan from inlined checklist to Copilot-assigned issue tasks #41715: [q] Shift Typist execution plan from inlined checklist to Copilot-assigned issue tasks → Closed — exploratory query, not a concrete change

---

Key Insights

• 🔑 Code blocks are the strongest signal: PRs with code blocks in the body have an 84% merge rate vs 69% without — a 15 percentage point gap. Including concrete code examples, diffs, or snippets strongly correlates with success.
• ⚠️ WIP tags are high-risk: [WIP]-prefixed PRs have only a 37% merge rate (15 merged, 26 closed out of 41). Avoid opening WIP PRs until the work is substantially complete.
• 📏 Longer bodies trend better, but specificity matters more: Long bodies (>1,000 chars) achieve 83% vs medium bodies (200–1,000 chars) at 75%. The key differentiator is concrete code, file references, and clear problem statements — not just length alone.

---

Recommendations

Based on today's analysis:

1. DO: Include a code block (snippet, diff, or config) in your PR body — this single factor raises the expected merge rate by ~15 points.
2. DO: Use precise, scoped action verbs in PR titles: fix(component):, bump X → Y, refactor: extract ... correlate with 82–90% success.
3. DO: Complete the work before opening a PR — avoid [WIP] tags; the 37% WIP merge rate is less than half the 81% baseline.
4. AVOID: Vague titles with words like proxy, path, phase 1, or wip — these correlate with closed PRs.
5. AVOID: Short PR bodies (<200 chars) without context — include a clear problem statement.

---

Historical Trends

Date	PRs	Merge Rate	Notable
2026-06-10	1000	79.6%	Baseline
2026-06-16	1000	80.3%	+0.7pp
2026-06-20	1000	79.5%	Minor dip
2026-06-23	1000	80.4%	Recovery
2026-06-26	1000	81.1%	Accelerating
2026-06-27	1000	81.2%	Peak ↑

Trend: Merge rate has steadily improved +1.6pp over the last 17 days. Week 26 (Jun 23–27) achieved the highest weekly rate at 84% — up from 78% in week 24.

📊 Raw weekly breakdown

Week	PRs	Merged	Closed	Rate
W23 (Jun 2–8)	135	111	24	82%
W24 (Jun 9–15)	329	258	71	78%
W25 (Jun 16–22)	251	204	47	81%
W26 (Jun 23–27)	285	236	45	84%

---

References: §28301554682

Generated by 🔍 Copilot PR Prompt Pattern Analysis · 76.7 AIC · ⌖ 22.6 AIC · ⊞ 7.5K · ◷

• expires on Jun 28, 2026, 1:07 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-28T21:07:08.769Z.

Closed by Workflow