[prompt-analysis] Copilot PR Prompt Analysis - 2026-10-29

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis - 2026-10-29

Summary

Analysis Period: Last 30 days
Total PRs: 612 | Merged: 503 (82.2%) | Closed: 98 (16.0%) | Open: 11 (1.8%)

Success rate among completed PRs: 83.7%. All PRs had non-empty bodies.

Prompt Categories and Success Rates

Categories are keyword-matched on PR titles (first match wins).

Category	Total	Merged	Success Rate (completed)
Bug Fix	69	59	86.8%
Feature Addition	90	74	84.1%
Refactoring	14	11	78.6%
Documentation	25	21	87.5%
Testing	7	5	71.4%
Other / uncategorized	407	333	83.3%

Prompt Analysis

✅ Successful Prompt Patterns

• Average body length (merged): 159 words
• Bodies referencing concrete file names (.go, .cjs, .md, .yml, .ts, .js): 78 merged vs 13 closed → 85.7% success, vs 83.3% without file references.
• Docs and bug-fix tasks scored highest.

Examples:

1. PR Guard Copilot sandbox startup against RUNNER_TEMP loss #72933: Guard Copilot sandbox startup against RUNNER_TEMP loss → Merged
2. PR docs: clarify workflow-specific Copilot model availability #72934: docs: clarify workflow-specific Copilot model availability → Merged
3. PR Link work queue footers to creation and completion revisions #72632: Link work queue footers to creation and completion revisions → Merged

❌ Unsuccessful Prompt Patterns

• Average body length (closed): 137 words (open: 108)
• Closed PRs concentrate in model/alias/config-behaviour changes and infra-specific fixes.

Examples:

1. PR Fix pi alias model paths on arc-dind #72937: Fix pi alias model paths on arc-dind → Closed
2. PR Honor disabled AI-credit budgets in runtime AWF config #72935: Honor disabled AI-credit budgets in runtime AWF config → Closed
3. PR Allow on.bots Apps to check out same-repository PR edits #72466: Allow on.bots Apps to check out same-repository PR edits → Closed
4. PR Exclude Copilot utility models from alias selection and restore detection fallbacks #72404: Exclude Copilot utility models from alias selection... → Closed

Key Insights

• Length: merged PR prompts are ~16% longer on average than closed ones (159 vs 137 words), suggesting more context helps.
• File references: modest lift (+2.4 pts) when specific files are named; effect is small, not strongly significant.
• Category: refactoring and testing samples are small (<15), so rates are not statistically reliable. Overall success is stable at ~83–85%.

Recommendations

1. DO: Include background and expected behaviour (longer, contextual prompts merge more).
2. DO: Name specific files/functions when known.
3. AVOID: Ambiguous policy/behaviour changes in model-alias and environment-specific areas without clear acceptance criteria; these account for many closures.

Historical Trends

Date	Merged	Closed	Success Rate
2026-10-29	503	98	83.7%
2026-10-25	506	89	85.0%
2026-10-22	165	29	85.1%
2026-10-20	513	101	83.5%

Trend: Stable success rate in the 83–85% range.

---

Generated by Copilot PR Prompt Analysis (Run: 40950858993)

Generated by 🔍 Copilot PR Prompt Pattern Analysis · copilot · auto · ~14 AIC · ⌖ ~1 AIC · ⊞ 11.6K · ◷

• expires on Oct 30, 2026, 12:43 PM UTC-08:00