[prompt-analysis] Copilot PR Prompt Analysis - 2026-07-13

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis — 2026-07-13

Summary

Analysis Period: Last 30 days
Total PRs: 1,000 | Merged: 780 (78.6%) | Closed: 212 (21.2%) | Open: 8 (0.8%)

---

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Refactor	25	21	4	84.0%
Bug Fix	781	615	158	79.6%
Documentation	58	45	13	77.6%
Testing	64	49	15	76.6%
Feature	58	42	16	72.4%
Other	14	8	6	57.1%

---

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: 183 words
• Most common keywords: workflow, actions, agent, spec, implement, fix
• WIP mentions: only 1.5% of merged PRs
• File references present: 95.4% of merged PRs
• Code blocks present: 82.4% of merged PRs

Example merged prompts

PR #45273 — pkg/stats: implement Sum, Variance, and StdDev

`pkg/stats/README.md` documented Sum(), Variance(), and StdDev() but none were implemented in statvar.go, causing an 80% spec quality score and test failures...

PR #45225 — Improve ExtractWorkflowNameFromFile hot-path parsing

BenchmarkExtractWorkflowNameFromFile regressed materially versus historical baseline despite low allocations. This change restores performance by...

PR #45224 — Restore docs home page skip-link target and prevent mobile overflow

The docs home page regressed in two user-visible ways on mobile: the rendered page no longer exposed the expected Starlight skip-link...

---

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: 181 words (nearly identical to merged)
• WIP mentions: 15.1% of closed PRs vs 1.5% merged — 10× higher
• Code blocks present: 70.8% vs 82.4% merged — notably lower
• Common closed PR keywords: wip, draft, general, batch, scaffold

Example closed prompts

PR #45110 — docs: prerender llms.txt and add contract coverage for AI discovery

The docs site GEO audit flagged https://github.github.com/gh-aw/llms.txt as missing, which is a high-weight discovery signal for AI crawlers...

PR #45106 — SPDD 2026-07-12: implement intent/authz enforcement, compliance

Addresses the SPDD daily plan covering 5 spec files (rotation indices 5–9). The critical P0 gap: AuthorizeTool policy...

Note: 4 closed PRs contained automated "agentic threat detected" caution blocks, triggering mandatory manual review that led to closure.

---

Key Insights

• WIP signals strongly predict closure: PRs mentioning "wip" or "draft" in their prompt are closed at a 10× higher rate (15.1% vs 1.5%). Avoid submitting prompts before the task is fully scoped.
• Code blocks correlate with merges: Including concrete code examples or diffs in the prompt body raised merge rates (+11.6 percentage points vs no code blocks). Prompts with \``` fences are more likely to get merged.
• Refactor prompts outperform feature prompts: Refactoring tasks (84.0%) consistently outperform new feature additions (72.4%) — likely because scope is more bounded and verifiable.
• Prompt length is not a differentiator: Merged (183 words) and closed (181 words) prompts are nearly identical in length — content quality matters more than verbosity.
• Agentic threat blocks cause closures: 4 PRs were closed after automated threat detection flagged content. Review prompt content before submission to avoid triggering security guardrails.

---

Recommendations

Based on today's analysis:

1. DO: Reference specific file paths and function names — 95.4% of merged PRs include file references.
2. DO: Include code blocks or diff excerpts in your prompt — 82.4% of merged PRs use them.
3. DO: Prefer targeted refactors over broad feature additions when possible — highest success rate (84%).
4. AVOID: Marking prompts as WIP or draft before submission — this is the single strongest predictor of closure (15× risk).
5. AVOID: Vague or scaffolding-style prompts (scaffold, general improvement) — these fall into the "other" category with only 57% success.

---

Historical Trends (Last 7 Days)

Date	PRs	Success Rate	Top Category
2026-07-13	1,000	78.6%	bug_fix
2026-07-12	1,000	78.9%	bug_fix
2026-07-11	1,000	79.1%	bug_fix
2026-07-10	1,000	79.9%	bug_fix
2026-07-08	1,000	80.3%	bug_fix
2026-07-07	1,000	80.0%	bug_fix
2026-07-06	1,000	80.4%	feature

Trend: ⚠️ Success rate has declined ~1.8 percentage points over the past week (80.4% → 78.6%). The shift may reflect more complex or speculative tasks being submitted. Monitor whether this stabilizes.

---

References: §29284344079

Generated by 🔍 Copilot PR Prompt Pattern Analysis · 33.7 AIC · ⌖ 11.1 AIC · ⊞ 7.4K · ◷

• expires on Jul 14, 2026, 1:03 PM UTC-08:00

[github-actions[bot]]

Smoke ping from run 29285543457 — discussion comment path is live. ✅

📰 BREAKING: Report filed by Smoke Copilot · 124.2 AIC · ⌖ 7.89 AIC · ⊞ 19K · ◷
_{Comment /smoke-copilot to run again
Add label smoke to run again}