[prompt-analysis] Copilot PR Prompt Analysis - February 7, 2026

[github-actions[bot]]

🤖 Copilot PR Prompt Pattern Analysis

Analysis Period: Last 30 days (January 8 - February 7, 2026)
Total PRs: 1,000 | Merged: 679 (68.7%) | Closed: 310 (31.3%) | Open: 11 (1.1%)

Prompt Categories and Success Rates

Category	Total	Merged	Success Rate
Other	13	10	76.9% ⭐
Refactor	152	110	72.4%
Documentation	552	378	68.5%
Feature Addition	702	480	68.4%
Testing	551	376	68.2%
Update	544	370	68.0%
Bug Fix	754	509	67.5%

Key Observation

Success rates are remarkably consistent across categories (67-69%), with refactoring PRs performing slightly better at 72.4%.

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Action verbs in titles: 72.9% of merged PRs vs 62.3% of closed
• File references: 88.7% of merged PRs mention specific files vs 81.6% of closed
• Average prompt length: 401 words (concise and focused)
• WIP/Draft tags: Only 4.1% of merged PRs marked as WIP

View Successful Examples

1. PR #14286: Fix invalid checkout-pr output references in workflows without contents permission

   • Why it succeeded: Clear action verb ("Fix"), specific file references, concrete problem statement
   • Body preview: "Two workflows (issue-triage-agent, weekly-issue-summary) reference steps.checkout-pr.outputs.checkout_pr_success but lack contents: read permi..."
   • Status: ✅ MERGED

2. PR #14285: docs: add missing hash-frontmatter command to CLI reference

   • Why it succeeded: Action verb ("add"), specific documentation gap identified, clear scope
   • Body preview: "The hash-frontmatter command was fully implemented and registered but undocumented, making it undiscoverable to users..."
   • Status: ✅ MERGED

3. PR #14275: Refactor threat detection tests: testify assertions, table-driven patterns, helper functions

   • Why it succeeded: Clear action ("Refactor"), specific file mentioned, detailed approach
   • Body preview: "Improves test quality in threat_detection_file_access_test.go by migrating to testify assertions, table-driven patterns..."
   • Status: ✅ MERGED

4. PR #14273: docs: add missing health command to CLI reference

   • Why it succeeded: Action verb ("add"), specific gap, concise scope
   • Body preview: "The health command was fully implemented and registered as an analysis command but absent from CLI documentation..."
   • Status: ✅ MERGED

5. PR #14260: Add fuzzy matching "did you mean" suggestions for engine and tool validation

   • Why it succeeded: Clear action ("Add"), specific feature, detailed context about problem being solved
   • Body preview: "Implementing improvements to enhance error message quality in the gh-aw compiler based on the Daily Syntax Erro..."
   • Status: ✅ MERGED

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• WIP/Draft tags: 33.2% of closed PRs marked as WIP vs only 4.1% of merged (8.1x more likely to be closed)
• Missing action verbs: 37.7% of closed PRs lack action verbs in titles vs 27.1% of merged
• File references: 81.6% mention files (still high, but lower than merged PRs)
• Average prompt length: 419 words (slightly longer, possibly less focused)

View Unsuccessful Examples

1. PR #14321: Require temporary_id for draft issue creation (breaking change)

   • Why it failed: Breaking change in title signals risk, no clear problem statement
   • Body preview: "Addressed all PR review feedback to improve code quality and efficiency..."
   • Status: ❌ CLOSED (not merged)

2. PR #14293: Fix update-project to deduplicate draft issues with temporary ID support

   • Why it failed: Complex multi-part fix, unclear scope
   • Body preview: "Fixes issue where update-project created duplicate draft issues when updating fields like status..."
   • Status: ❌ CLOSED (not merged)

3. PR #14265: [WIP] Revert gh-aw-mcpg version in constants.go

   • Why it failed: Explicitly marked as WIP, no clear completion criteria
   • Body preview: "Thanks for asking me to work on this. I will get started on it and keep this PR's description up to date..."
   • Status: ❌ CLOSED (not merged)

4. PR #14305: Document custom job dependencies (needs) for agentic workflows

   • Why it failed: Missing action verb in title ("Document" implied but not stated clearly)
   • Body preview: "Investigation Complete: Matrix Jobs and Job Dependencies for Project Boards and Orchestration..."
   • Status: ❌ CLOSED

5. PR #14294: Analysis: ProjectOps + Orchestration pain points with minimal architectural solutions

   • Why it failed: "Analysis" in title suggests investigation, not implementation; unclear deliverable
   • Body preview: "Orchestration patterns (dispatcher/worker/aggregator) combined with projectOps safe outputs exhibit four pain points..."
   • Status: ❌ CLOSED

Key Insights

🎯 Critical Success Factors

1. WIP/Draft Status is a Red Flag

   • WIP/Draft PRs are 8.1x more likely to be closed than merged
   • Only 4.1% of merged PRs are marked WIP vs 33.2% of closed PRs
   • Recommendation: Avoid opening PRs until they're ready for review

2. Action Verbs Drive Success

   • PRs with action verbs in titles have 17% higher merge rate
   • Top action verbs: "Fix", "Add", "Update", "Refactor", "Implement"
   • Missing action verbs correlate with 37.7% of closed PRs

3. Specificity Matters

   • File references are 7 percentage points higher in merged PRs (88.7% vs 81.6%)
   • Merged PRs tend to be slightly more concise (401 vs 419 words)
   • Successful prompts clearly state what file/function will change

4. Category Performance is Consistent

   • All major categories maintain 67-72% success rates
   • Refactoring PRs perform best (72.4%), possibly due to clearer scope
   • Bug fixes have lowest rate (67.5%), likely due to complexity

Recommendations

✅ DO: Follow These Patterns

1. Start titles with action verbs

   ✅ "Fix invalid checkout-pr output references"
   ✅ "Add fuzzy matching suggestions for validation"
   ❌ "Invalid output references issue"
   ❌ "Analysis: ProjectOps pain points"
   ```
   
   

2. Reference specific files/components

   ✅ "Refactor threat_detection_file_access_test.go with testify"
   ❌ "Improve test quality in project"
   

3. Keep prompts focused and concise

   • Target 300-500 words for body
   • State problem clearly upfront
   • Provide concrete scope

4. Avoid WIP/Draft markers

   • Only open PRs when ready for review
   • If you must use WIP, have clear completion criteria

❌ AVOID: Anti-Patterns

1. Investigation/Analysis titles without clear deliverables

   • "Analysis: ..." → Be specific about what will be implemented
   • "Document..." → Use "Add documentation for..."

2. Breaking changes without strong justification

   • Signal risk upfront
   • Provide migration path
   • Get buy-in before implementation

3. Vague or compound problems

   • "Fix X and Y and add Z" → Split into focused PRs
   • "Improve quality" → Specify what quality metric

Historical Trends (2-Day Comparison)

Date	PRs	Merged	Success Rate	Top Category
Feb 7, 2026	1,000	679	68.7%	Other (76.9%)
Feb 6, 2026	1,000	687	69.0%	Update (70.9%)

Trend: Success rates remain remarkably stable day-over-day (68-69%). The key differentiator is not category choice but prompt clarity and WIP status.

Prompt Quality Checklist

Before creating a Copilot PR, ensure your prompt:

• Starts with a clear action verb (Fix, Add, Update, etc.)
• References specific files or components
• States the problem or goal concisely (1-2 sentences)
• Provides context but stays focused (~300-500 words)
• Is NOT marked as WIP or Draft
• Has a single, clear scope (not multiple unrelated changes)

---

References:

• §21779962052

Historical data stored in cache memory at /tmp/gh-aw/cache-memory/prompt-analysis/history.json for trend tracking.

AI generated by Copilot PR Prompt Pattern Analysis

• expires on Feb 14, 2026, 12:21 PM UTC

[github-actions[bot]]

💥 WHOOSH! KAPOW! 💥

The Smoke Test Agent just BLAZED through here at lightning speed! ⚡️🔥

🦸 Agent Status: OPERATIONAL
🎯 Mission: Engine validation COMPLETE
🚀 Next Stop: Saving the world, one test at a time!

THWACK! All systems GO! BOOM!

🦾 Powered by: Claude Sonnet 4.5
📍 HQ: §21780092720

Stay awesome! 💪✨

AI generated by Smoke Claude

[github-actions[bot]]

🔮 The ancient spirits stir; the smoke test agent has walked these halls and left its sigil in the margins. The omens are clear, and the circuits hum with quiet approval.

AI generated by Smoke Codex

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-14T12:21:33.831Z.

Closed by Workflow