[prompt-analysis] Copilot PR Prompt Analysis - 2026-04-25

[github-actions[bot]]

Summary

Analysis Period: Last 30 days | Total PRs: 1,000 | Merged: 780 (78.1%) | Closed: 218 (21.8%) | Open: 2 (0.2%)

Overall merge rate is strong at 78.1%, a slight drop from yesterday's 78.1% (vs 78.8% on 2026-04-23 and 80.2% on 2026-04-17).

---

Prompt Categories and Success Rates

Category	Total	Merged	Success Rate
Bug Fix	889	699	79%
Feature Addition	94	67	71%
Documentation	5	4	80%
Testing	5	4	80%
Refactoring	2	2	100%

Note: Categories are not mutually exclusive; PRs are assigned to the first matching category based on title/body keywords.

---

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt body length: ~1,081 words
• Most common keywords: workflow, github, changes, agent, workflows, safe, copilot, actions, added, test, step, engine, file, output, setup
• Titles are specific and scoped (e.g., include the component name and symptom)

Example merged PRs:

1. PR Add Kreuzberg document intelligence MCP shared workflow #28392: "Add Kreuzberg document intelligence MCP shared workflow" — Adds a concrete, named component with clear purpose → Merged
2. PR fix: TypeScript type errors in push_to_pull_request_branch cross-repo support #28388: "fix: TypeScript type errors in push_to_pull_request_branch cross-repo support" — Precise component path + error type in title → Merged
3. PR fix: remove run_id from trending cache key and enforce via compiler validation #28387: "fix: remove run_id from trending cache key and enforce via compiler validation" — Root cause identified in title, fix scoped to specific mechanism → Merged

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt body length: ~1,042 words (slightly shorter)
• Most common keywords: workflow, github, copilot, workflows, changes, agent, lock, make, codex, safe, actions, validation, step, issue, update
• Presence of [WIP] prefix strongly correlates with closure
• Prompts mentioning codex and lock files appear more often in closed PRs (potential compilation/recompile failures)

Example closed PRs:

1. PR [WIP] Add tests for cross-repo push_to_pull_request_branch #28380: "[WIP] Add tests for cross-repo push_to_pull_request_branch" — Work-in-progress with incomplete checklist → Closed
2. PR disable requireCleanGit for skill-optimizer CI runs #28378: "disable requireCleanGit for skill-optimizer CI runs" — Narrow workaround rather than root fix → Closed
3. PR [WIP] Fix failing GitHub Actions job deploy #28335: "[WIP] Fix failing GitHub Actions job deploy" — Vague title + agent exploratory plan in body → Closed

---

Key Insights

• [WIP] = risk signal: PRs with [WIP] in the title are disproportionately represented in closed PRs. Prompts that produce exploratory or incomplete work tend to be abandoned.
• Bug fix prompts dominate (89%) and perform well (79%): This repo primarily uses Copilot for targeted fixes. Success is highest when the prompt describes a concrete, reproducible symptom and points to specific code locations.
• Feature prompts underperform bug fixes (71% vs 79%): Feature prompts are less scoped. Adding context about the expected behavior and relevant files improves success rates.
• Prompt length barely differs between merged and closed (~1,081 vs ~1,042 words), suggesting length alone is not a predictor of success — specificity matters more.

---

Recommendations

1. DO: Include the specific file path, function name, or component in the prompt title — e.g., fix: <symptom> in <module>.
2. DO: For bug fixes, describe the root cause (not just the symptom) in the prompt body when known.
3. AVOID: Starting with [WIP] or asking Copilot to "explore and figure out a plan" — these produce tentative PRs that are usually closed.
4. DO: For feature additions, specify the expected interface/behavior and reference any related files or issues.
5. AVOID: Vague action verbs like "improve" or "update" without a concrete target — these correlate with closed PRs.

---

Historical Trends (Last 7 Days)

Date	PRs	Merged	Success Rate
2026-04-25	1,000	780	78.1%
2026-04-24	1,000	781	78.1%
2026-04-23	1,000	788	78.8%
2026-04-17	1,000	800	80.2%

Trend: Merge rate has declined ~2 percentage points over the past week (80.2% → 78.1%). The increase in closed PRs may reflect more exploratory/WIP Copilot usage or harder tasks being attempted.

---

References:

• §24923096887

Generated by Copilot PR Prompt Pattern Analysis · ● 610.1K · ◷

• expires on Apr 26, 2026, 5:03 AM UTC

[github-actions[bot]]

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

A newer discussion is available at Discussion #28551.