[prompt-analysis] Copilot PR Prompt Analysis - 2026-10-20

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-10-20

Summary

Analysis Period: Last 30 days
Total PRs: 622 | Merged: 513 (82.5%) | Closed: 101 (16.2%) | Open: 8

Merge rate among completed PRs: 83.5%. No PRs had an empty body.

Prompt Categories and Success Rates

Categories are keyword-matched on PR titles. "Other" are titles with no matching verb.

Category	Total (completed)	Merged	Success Rate
Bug Fix	74	64	86.5%
Feature Addition	92	73	79.3%
Refactoring	15	12	80.0%
Documentation	31	25	80.6%
Testing	16	13	81.3%
Other	386	326	84.5%

Prompt Analysis

✅ Successful Prompt Patterns

• Average prompt length: 170.8 words (median 105)
• Files or paths referenced in the body: 18.7% of merged PRs
• Narrow, single-purpose titles, for example:
  1. PR Record failed model.resolution for Pi sub-agents whose model fails the /reflect check #70385: Record failed model.resolution for Pi sub-agents whose model fails the /reflect check → Merged
  2. PR Fail activation early for stale provenance workflows #70348: Fail activation early for stale provenance workflows → Merged
  3. PR Record authoritative zero AIC evidence for pre-flight noops #70344: Record authoritative zero AIC evidence for pre-flight noops → Merged

❌ Unsuccessful Prompt Patterns

• Average prompt length: 198.2 words (median 102)
• Files or paths referenced in the body: 17.8% of closed PRs
• Closed PRs skew toward broad, new-capability work, for example:
  1. PR Support team-configurable default models #70349: Support team-configurable default models → Closed
  2. PR Generate GitHub-native workflow lockfiles during compilation #70345: Generate GitHub-native workflow lockfiles during compilation → Closed
  3. PR Use artifact-backed credit accounting for daily code metrics #69988: Use artifact-backed credit accounting for daily code metrics → Closed

Key Insights

• Bug fixes merge most often (86.5%). Feature additions are lowest (79.3%).
• Longer bodies correlate slightly with closure. Closed PRs average about 27 more words than merged ones, but the medians are nearly identical (105 vs 102). A few very long prompts drive the gap.
• File references don't separate outcomes. The rates are 18.7% vs 17.8%, so this is not a meaningful signal in this sample.
• Small samples: refactoring (15) and testing (16) are too small for firm conclusions.

Recommendations

1. DO: Scope prompts to one observable behavior, as in the merged examples above.
2. DO: Split large feature requests, such as multi-component compiler or config changes, into smaller tasks.
3. AVOID: Very long, multi-goal prompts for new capabilities. Their closure rate is higher.

Historical Trends

Date	PRs	Success Rate	Notes
2026-10-20	622	83.5%	Bug fix highest at 86.5%
2026-10-19	611	83.6%	Bug fix 88.3%
2026-10-15	n/a	84.2%	From repo-memory history
2026-10-12	n/a	84.0%	From repo-memory history

Trend: The merge rate is stable at about 84% across the last week.

---

Generated by Copilot PR Prompt Analysis (Run: 39494406496)

Generated by 🔍 Copilot PR Prompt Pattern Analysis · copilot · auto · 16.4 AIC · ⌖ 0.716 AIC · ⊞ 11.5K · ◷

• expires on Oct 21, 2026, 12:48 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Copilot PR Prompt Pattern Analysis.

A newer discussion is available at Discussion #70949.