[prompt-analysis] Copilot PR Prompt Analysis - 2026-06-16

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-06-16

Summary

Analysis Period: Last 30 days (2026-05-28 → 2026-06-16)
Total PRs: 1,000 | Merged: 797 (79.7%) | Closed: 196 (19.6%) | Open: 7 (0.7%)

Today (Jun 16): 33 new PRs with 92% success rate — highest single-day rate in recent history

Prompt Categories and Success Rates

Category	Total	Merged	Success Rate
Documentation	160	136	86%
Testing	193	162	84%
Refactoring	36	30	83%
Other	39	32	82%
Feature Addition	57	45	80%
Chore/Update	53	42	79%
Bug Fix	462	350	77%

Bug Fix has the lowest success rate despite being the most common category (46% of all PRs).

Prompt Analysis

Successful Prompt Patterns

Common characteristics in merged PRs:

• Average body length: 192 words (concise and focused)
• Start with the specific problem: inline code reference in first sentence (97% of merged PRs use backtick code refs)
• Common action verbs: fix, ensure, propagate, sync, add, consolidate, resolve, stop
• Root-cause statement before solution description
• 64% include specific file references

Example successful prompts:

1. PR #39653: "Threat-detection runs inherited Copilot BYOK env, but their AWF allowlist did not include the BYOK provider host. As a result, workflows that worked in the main agent job failed in the detection job..." → Merged

2. PR #39651: "Workflows with inline blocks could compile without the activation step when GitHub-context templating was suppressed..." → Merged

3. PR #39637: "WorkflowListItem and WorkflowStatus in pkg/cli duplicated 5 fields with identical tags. WorkflowStatus now embeds WorkflowListItem as the single source of truth..." → Merged

Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average body length: 218 words (13% longer — verbosity without clarity)
• 17% of closed PRs carry [WIP] markers (vs only 2% of merged)
• 29% of closed PRs were superseded by a newer PR
• 11% of closed PRs began with incomplete agent initialization phrases ("Thanks for asking", "I will get started")
• Analytics/report workflows: 48 closed (often replaced by improved versions)

Example unsuccessful prompts:

1. PR #39638: "Quick Start was Copilot-first and omitted two Claude-critical onboarding details: CLAUDE_CODE_OAUTH_TOKEN is unsupported..." → Closed (superseded)

2. PR #38156: "[WIP] Fix Daily Agent of the Day Blog Writer workflow failure due to AI credits — Thanks for asking me to work on this. I will get started on it..." → Closed (WIP, incomplete)

Key Insights

• Conciseness wins: Merged PR bodies average 192 words vs 218 for closed. Focused, specific prompts outperform verbose ones — trim background and get to the root cause.
• WIP is a red flag: PRs tagged as WIP are 8x more likely to be closed without merge. Avoid submitting PRs before the implementation is complete, or use draft PRs instead.
• Superseded churn accounts for 29% of closures: A significant portion of closed PRs are duplicate attempts. Check for recent related PRs before creating a new one to avoid redundant work.

Recommendations

Based on today's analysis:

1. DO: Start the PR body with a single-sentence root-cause statement using inline code references (backticks) for symbols and file paths — this pattern appears in 97% of merged PRs.
2. DO: Keep prompts under 200 words. Include: problem → impact → fix summary. Omit background that the reviewer can infer.
3. AVOID: Submitting work-in-progress PRs (WIP tag). They close at 8x the normal rate. Use GitHub draft PRs to signal in-progress work without the WIP-close risk.
4. AVOID: Creating new PRs without first searching for recent duplicates — 29% of closures are superseded work.

Historical Trends (Last 7 Days)

Date	PRs Created	Daily Merge Rate	Top Category
2026-06-16 (today)	33	92%	Bug Fix
2026-06-13	41	88%	Bug Fix
2026-06-12	31	87%	Bug Fix
2026-06-10	55	80%	Bug Fix
2026-06-09	66	82%	Testing
2026-06-08	72	74%	Bug Fix
2026-06-07	95	82%	Feature

Trend: Strong upward trajectory from 74% (Jun 8) to 92% (Jun 16). 30-day aggregate steady at ~80%. Documentation continues to lead all categories at 86% success rate.

Methodology

• Data: Top 1,000 Copilot PRs (copilot/* branches) from last 30 days in github/gh-aw
• Categories derived from keyword matching on PR title + body
• "Success rate" = merged / (merged + closed); open PRs excluded from rate calculation
• Body length measured in words on the full PR body field
• Historical trend uses daily PR creation counts from the rolling 30-day window

---

References: §27649686535

Generated by 🔍 Copilot PR Prompt Pattern Analysis · ◷

• expires on Jun 17, 2026, 1:44 PM UTC-08:00