[prompt-analysis] Copilot PR Prompt Analysis - 2026-10-01

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-10-01

Summary

Analysis Period: Last 30 days
Total PRs: 566 | Merged: 435 (76.9%) | Closed: 127 (22.4%) | Open: 4 (0.7%)

Merge rate among completed PRs: 77.4%.

Prompt Categories and Success Rates

Categories are assigned by title keyword matching (first match wins).

Category	Total (completed)	Merged	Success Rate
Bug Fix	136	118	86.8%
Feature Addition	110	70	63.6%
Refactoring	11	10	90.9%
Documentation	17	11	64.7%
Testing	14	10	71.4%
Other	274	216	78.8%

Prompt Analysis

✅ Successful Prompt Patterns

• Average prompt (PR body) length: 186 words for merged PRs
• Bug fixes and refactors with a narrow, well-defined scope merge most often
• Prompts without explicit file or path references merged at 86% (300 of 349), versus 63% (135 of 213) for prompts with file references

Example merged PRs:

1. Add built-in log, set, map, table, and counter ledgers #64721: Add built-in log, set, map, table, and counter ledgers → Merged
2. Consolidate ledger compaction into two Agentic Maintenance jobs #64709: Consolidate ledger compaction into two Agentic Maintenance jobs → Merged
3. Move ledger compaction into Agentic Maintenance (untrusted plan job / trusted apply job) #64599: Move ledger compaction into Agentic Maintenance → Merged

❌ Unsuccessful Prompt Patterns

• Average prompt length: 361 words, about twice as long as for merged PRs
• Feature additions have the highest closure rate (36%)
• Exploratory, "investigate", or multi-part design tasks are often closed

Example closed PRs:

1. Add ledger compaction transactions to safe outputs and replay #64585: Add ledger compaction transactions to safe outputs and replay → Closed
2. Investigate campaign-query memory estimation gap #64583: Investigate campaign-query memory estimation gap → Closed
3. Add declarative ledger types for collection replay and latest-wins tables #64581: Add declarative ledger types for collection replay and latest-wins tables → Closed

Key Insights

• Length matters: closed PRs have prompts about 2x longer, which suggests oversized or multi-goal tasks.
• Features are riskiest: 63.6% success vs 86.8% for bug fixes.
• File references are not a success signal here: PRs mentioning files are more likely to be larger or more complex, so this is likely confounded by task size.
• Closed PRs cluster on iterative design work such as the ledger series (Add declarative ledger types for collection replay and latest-wins tables #64581, Add ledger compaction transactions to safe outputs and replay #64585), where later PRs superseded earlier ones.

Recommendations

1. DO: Keep prompts focused on a single, bounded outcome, ideally under about 200 words.
2. DO: Split large features into incremental PRs, and include a reproduction or error message for bug fixes.
3. AVOID: Open-ended "investigate" tasks and multi-part design changes in one prompt.

Historical Trends

Date	PRs	Success Rate	Top Category
2026-10-01	566	77.4%	Bug Fix (86.8%)
2026-09-30	560	77.5%	Bug Fix (84.2%)

Trend: Stable overall merge rate, with feature success holding around 64%.

---

Generated by Copilot PR Prompt Analysis (Run: 36923844708)

Generated by 🔍 Copilot PR Prompt Pattern Analysis · copilot · auto · 14.1 AIC · ⌖ 6.14 AIC · ⊞ 10.8K · ◷

• expires on Oct 2, 2026, 12:53 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 #65102.