[prompt-analysis] Copilot PR Prompt Analysis - 2026-06-17

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis — 2026-06-17

Analysis Period: Last 30 days
Total PRs: 1,000 | Merged: 804 (80.7%) | Closed (not merged): 192 (19.3%) | Open: 4

---

Prompt Categories and Success Rates

Category	Total	Merged	Success Rate	Trend
Refactoring	14	13	93%	⬆️
Other/Misc	13	12	92%	—
Documentation	41	36	88%	➡️
Testing	228	193	85%	➡️
Bug Fix	671	525	79%	➡️
Feature Addition	33	25	76%	⬇️

Note: Categories are assigned by keyword matching on title + body.

---

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: 172 words
• Top action verbs: add (19%), fix (17%), remove (6%), update (4%)
• 82.8% of merged PRs include code blocks (vs 71.3% without — +11.5pp advantage)
• 72.3% reference specific filenames in backtick notation
• Conventional commit prefix in title: 83.2% success rate

Example successful prompts:

1. PR #39847: fix: retry defender scan on transient WinDefend service unavailability — clear error context (hr = 0x800106ba), specific file refs (cgo.yml), 110 words → Merged
2. PR #39875: Suppress ANSI installer colors in CI and explicit no-color environments — concrete problem statement, specific env conditions, bulleted changes → Merged
3. PR #39713: fix: enforce minLength on update_release body to block placeholder submissions — references specific run ID, root cause identified → Merged

---

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: 189 words (longer than merged — verbosity can signal scope creep)
• Top action verbs: fix (23.4%), add (14.6%), refactor (6.2%)
• [WIP] tag in title: only 35% merge rate (vs 80.7% overall)
• Keywords container, mcp, outputs, comments over-represented in closed PRs

Example unsuccessful prompts:

1. PR #39695: [WIP] Fix safe outputs MCP container path for consumer repos — vague WIP status, large infrastructure scope → Closed
2. PR #39009: [WIP] Refactor to replace map[string]bool with map[string]struct{} — mechanical refactor without behavioral benefit, WIP tag → Closed
3. PR #39725: Implement matrix support and shared workflow category merging in compiler — very broad scope, generic AI Agent body → Closed

---

Key Insights

• 📦 Code blocks matter: PRs with code blocks succeed at 82.8% vs 71.3% without — always include relevant snippet context (+11.5pp).
• 🏷️ WIP signals trouble: [WIP] PRs succeed at only 35% — 2.3x more likely to be closed than the overall rate.
• 🎯 Scoped refactoring wins: Well-described, bounded refactoring PRs achieved 93% merge rate — the key is clear rationale and defined scope.
• 📝 Short & specific outperforms verbose: 20–49 word prompts achieve 89% success; 200+ words drops to 81%.

---

Recommendations

1. DO: Include code blocks or backtick-quoted filenames — they signal specificity and yield a +11.5pp success advantage.
2. DO: Use conventional commit prefixes (fix:, chore:, docs:, feat:) — they correlate with 83.2% merge rate.
3. DO: For refactoring tasks, define exact scope and rationale — scoped refactors reach 93%.
4. AVOID: Opening PRs with [WIP] in the title unless iteration is expected — merge rate drops to 35%.
5. AVOID: Overly broad prompts touching MCP/container/agent infrastructure simultaneously — these correlate with closed PRs.

---

Historical Trends (Last 6 Days)

Date	PRs	Merged	Success Rate	Key Insight
2026-06-17	1,000	804	80.7%	Refactoring 93%; code blocks +11.5pp
2026-06-16	1,000	797	80.0%	Doc prompts led at 86%; WIP 8x close rate
2026-06-15	1,000	792	79.4%	optimize/handle/move verbs at 100% rate
2026-06-14	1,000	791	79.1%	update/remove verbs at 92%+
2026-06-13	1,000	792	80.0%	Stable
2026-06-12	1,000	786	79.0%	Slight dip

Trend: Overall success rate has climbed from 79.0% to 80.7% over the past week — a consistent improvement. Refactoring and documentation prompts continue to outperform bug fix prompts.

---

References:
\u00a727720481724

Generated by 🔍 Copilot PR Prompt Pattern Analysis · ◷

• expires on Jun 18, 2026, 1:28 PM UTC-08:00