[prompt-analysis] Copilot PR Prompt Analysis - 2026-07-09

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis — 2026-07-09

Summary

Analysis Period: Last 30 days
Total PRs: 1000 | Merged: 792 (80.8%) | Closed: 195 (19.9%) | Open: 13 (1.3%)

---

Prompt Categories and Success Rates

Category	Total	Merged	Success Rate
Bug Fix	824	655	80.8%
Testing	109	87	79.8%
Feature Addition	21	17	81.0%
Documentation	22	17	77.3%
Refactoring	13	9	69.2%
Other	11	7	63.6%

---

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: 181.8 words
• Bullet list present: 97.2% of merged PRs
• Code blocks present: 82.2% of merged PRs
• File references present: 71.1% of merged PRs
• WIP/Draft label: only 1.3% of merged PRs
• Common keywords: call, type, string, comment, summary, step

Example successful prompts

1. PR compiler: add evals job for BinEval binary evaluations #43700 — compiler: add evals job for BinEval binary evaluations → Merged
2. PR fix(security): replace curl|sh installer pipes with download+verify (RGS-018) #43716 — fix(security): replace curl|sh installer pipes with download+verify (RGS-018) → Merged
3. PR Fix external detection job setup for codex and pi workflows #40954 — Fix external detection job setup for codex and pi workflows → Merged

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: 180.5 words (virtually identical to merged)
• Bullet list present: 92.3% (vs 97.2% for merged)
• Code blocks present: 71.3% (vs 82.2% for merged — 10.9 pp gap)
• File references present: 66.7% (vs 71.1% for merged)
• WIP/Draft label: 13.8% — 10× higher than merged

Example unsuccessful prompts

1. PR fix(claude): inject COPILOT_GITHUB_TOKEN, scope ANTHROPIC_BASE_URL to Anthropic provider, fix provider-aware api-proxy reflect check, convert Copilot env vars to Anthropic vars in harness, fix smoke model for reasoning effort, and mark model-provider: ... #40930 — Multi-fix kitchen-sink prompt → Closed
2. PR [WIP] Analyze daily breaking changes for 2026-07-06 #43822 — [WIP] Analyze daily breaking changes for 2026-07-06 → Closed
3. PR [WIP] Expand personas and add systematic cache-memory rotation #42090 — [WIP] Expand personas and add systematic cache-memory rotation → Closed

---

Key Insights

• WIP/Draft labels are a strong failure signal: Closed PRs carry WIP/Draft labels at 13.8% vs only 1.3% for merged — a 10× gap.
• Code blocks correlate with success: Merged PR bodies include code blocks in 82.2% vs 71.3% for closed (10.9 pp gap).
• Prompt length does not matter — structure does: Merged and closed PRs have nearly identical average word counts (~181 words), but merged PRs use bullet lists (97.2%) and file references (71.1%) significantly more.

---

Recommendations

1. DO: Include specific file paths and function names — file references appear in 71.1% of merged PRs.
2. DO: Use bullet lists to enumerate changes — 97.2% of merged PRs have structured lists.
3. DO: Include code blocks with error messages or before/after snippets — present in 82.2% of merged PRs.
4. AVOID: WIP/Draft labels — 13.8% of closed PRs carry these vs 1.3% of merged.
5. AVOID: Cramming multiple unrelated fixes into a single PR/prompt.

---

Historical Trends (last 7 days)

Date	PRs	Merged	Success Rate
2026-07-09	1000	792	80.8%
2026-07-08	1000	795	80.2%
2026-07-07	1000	792	80.0%
2026-07-06	1000	793	80.4%
2026-07-05	1000	805	80.8%
2026-07-04	1000	798	81.1%
2026-07-03	1000	801	81.8%

Trend: Success rates are stable at 80–82%. Bug Fix dominates daily volume. Code block presence and WIP rate are consistent structural signals across all 7 days.

---

References:

• §29050184752

Generated by 🔍 Copilot PR Prompt Pattern Analysis · 227.6 AIC · ⌖ 23 AIC · ⊞ 7.4K · ◷

• expires on Jul 10, 2026, 1:22 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-07-10T21:22:21.082Z.

Closed by Workflow