[prompt-analysis] Copilot PR Prompt Pattern Analysis - January 25, 2026

[github-actions[bot]]

📊 Summary

Analysis Period: Last 30 days (1,000 PRs analyzed)
Total PRs: 1000 | Merged: 644 (64.7%) | Closed: 352 (35.3%) | Open: 4

📈 Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Refactor	2	2	0	100.0%
Feature	197	141	56	71.6%
Documentation	21	14	7	66.7%
Testing	109	72	37	66.1%
Bug Fix	619	389	230	62.8%
Other	48	26	22	54.2%

🎯 Key Findings

Prompt Length Analysis

• Merged PRs: Average of 120 words per prompt
• Closed PRs: Average of 144 words per prompt
• Insight: Closed PRs tend to have longer prompts (+24 words), suggesting verbosity may correlate with rejection

Category Insights

1. Refactoring shows 100% success rate (small sample: 2 PRs)
2. Feature additions perform better than expected at 71.6% success
3. Testing improvements maintain solid 66.1% success rate
4. Bug fixes are the most common category (619 PRs) with 62.8% success
5. Other/Uncategorized prompts have lowest success at 54.2%

✅ Successful Prompt Patterns

Most common keywords in merged PRs:

• workflow (1183 occurrences)
• github (852 occurrences)
• workflows (546 occurrences)
• changes (517 occurrences)
• agent (481 occurrences)
• githubnext (425 occurrences)
• actions (409 occurrences)
• campaign (368 occurrences)
• run (346 occurrences)
• mcp (343 occurrences)

Example successful prompts:

PR #11730: Add missing MCP properties to main workflow schema

Category: testing
Word count: 257
Outcome: ✅ Merged

[plan] Add missing MCP properties to main workflow schema

## Objective
> 
> Add 4 missing MCP properties to `main_workflow_schema.json` that are already implemented in the codebase but not validated by the schema.
> 
> ## Context
> 
> From discussion githubnext/gh-aw#11706: The main schema is missing `registry`, `allowed`, `entrypoint`, and `mounts` properties that are already implemented in `pkg/parser/mcp.go`, `pkg/workflow/mcp-config-custom.go`, and tested in `mcp_config_compilation_test.go`...
``````

[View PR](https://github.com/githubnext/gh-aw/pull/11730)

</details>

<details>
<summary><b>PR #11719: Remove `mode` field from custom MCP servers schema and add migration codemod</b></summary>

**Category**: feature  
**Word count**: 272  
**Outcome**: ✅ Merged

``````
[plan] Clarify or deprecate "mode" property in main schema

## Objective
> 
> Determine the purpose of the `mode` property in main schema and either document it clearly or deprecate it.
> 
> ## Context
> 
> From discussion githubnext/gh-aw#11706: The main schema defines both `mode` and `type` with identical enum values, but MCP config schema only has `type`. It's unclear if `mode` is:
> - An alias for `type`
> - A deprecated field
> - A workflow-specific property
> 
> ## Approach
> 
> 1. Search ...
``````

[View PR](https://github.com/githubnext/gh-aw/pull/11719)

</details>

### ❌ Unsuccessful Prompt Patterns

**Most common keywords in closed PRs:**
- `workflow` (808 occurrences)
- `github` (664 occurrences)
- `githubnext` (443 occurrences)
- `agent` (424 occurrences)
- `actions` (418 occurrences)
- `run` (416 occurrences)
- `workflows` (361 occurrences)
- `campaign` (353 occurrences)
- `runs` (345 occurrences)
- `agentic` (339 occurrences)

**Example unsuccessful prompts:**

<details>
<summary><b>PR #11745: Add CLI command for campaign project creation with views and custom fields</b></summary>

**Category**: feature  
**Word count**: 62  
**Outcome**: ❌ Closed

``````
Campaign project creation was only accessible through the automated workflow (applying `create-agentic-campaign` label). This adds direct CLI access.

## Changes

**New command: `gh aw campaign create-project`**
- Creates GitHub Projects V2 via GraphQL API
- Configures views (board/table/roadmap) via REST API
- Defines custom fields (TEXT, DATE, SINGLE_SELECT, NUMBER)
- Returns project URL for campaign spec integration

**Implementation:**
- `pkg/cli/campaign_create_project_command.go` - Command...
``````

[View PR](https://github.com/githubnext/gh-aw/pull/11745)

</details>

<details>
<summary><b>PR #11727: Align MCP schema validation constraints between main and config schemas</b></summary>

**Category**: bug_fix  
**Word count**: 231  
**Outcome**: ❌ Closed

``````
[plan] Align pattern constraints between main and MCP config schemas

## Objective
> 
> Fix pattern constraint mismatches between `main_workflow_schema.json` and `mcp_config_schema.json` to ensure consistent validation.
> 
> ## Context
> 
> From discussion githubnext/gh-aw#11706: The schemas have conflicting pattern constraints that cause inconsistent validation:
> - Environment variable patterns differ (lowercase vs uppercase-only)
> - Container image pattern missing in main schema
> - URL/Comm...

View PR

💡 Key Insights

Based on analysis of 996 completed PRs (merged or closed):

1. Conciseness Matters: Merged PRs average 120 words vs 144 words for closed PRs. Shorter, focused prompts correlate with higher success rates.

2. Category Performance: Feature additions (71.6%) outperform bug fixes (62.8%), suggesting Copilot handles new functionality better than fixing existing issues.

3. Common Ground: Both merged and closed PRs share similar technical vocabulary (workflow, github, actions), indicating keyword choice alone doesn't determine success.

4. Success Rate: Overall 64.7% merge rate shows Copilot-generated PRs have better-than-chance success, but there's room for improvement.

🎯 Recommendations

Based on today's analysis:

1. DO Keep prompts concise: Aim for 100-150 words. Provide necessary context but avoid excessive detail.

2. DO Focus on feature additions: When possible, frame work as adding new functionality rather than fixing existing code.

3. DO Provide clear objectives: Successful prompts clearly state what needs to be done and why.

4. AVOID Overly long explanations: Prompts exceeding 200 words show diminishing returns.

5. AVOID Vague categorization: The "other" category has lowest success at 54.2% - be specific about what you're asking for.

📝 Methodology

This analysis:

• Examined all Copilot-generated PRs from the last 30 days
• Extracted prompts from PR bodies using pattern matching
• Categorized prompts by intent (bug fix, feature, testing, etc.)
• Calculated success rates by category and analyzed prompt characteristics
• Compared merged vs closed PRs to identify correlation patterns

References:

• Workflow Run §21332196111

AI generated by Copilot PR Prompt Pattern Analysis

• expires on Feb 1, 2026, 12:03 PM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-01T12:03:07.780Z.