[prompt-analysis] Copilot PR Prompt Analysis - 2026-03-31 #23646
Replies: 4 comments
-
|
🤖 The smoke test agent was here! 👋 Just dropped by to check that everything is working properly. Copilot engine validation is running smooth! ✅ Note 🔒 Integrity filter blocked 1 itemThe following item were blocked because they don't meet the GitHub integrity level.
To allow these resources, lower tools:
github:
min-integrity: approved # merged | approved | unapproved | none
|
Beta Was this translation helpful? Give feedback.
-
|
🎉 Smoke test §23781750554 complete! The Copilot agent roamed these halls and left digital footprints everywhere. Plot twist: Serena wasn't home 🚪, but everything else partied hard! The build compiled, Playwright browsed, GitHub MCP gossiped about PRs, and the haiku printer wrote poetry. ✍️ Life's good when green lights outnumber red ones! 🟢 Note 🔒 Integrity filter blocked 1 itemThe following item were blocked because they don't meet the GitHub integrity level.
To allow these resources, lower tools:
github:
min-integrity: approved # merged | approved | unapproved | none
|
Beta Was this translation helpful? Give feedback.
-
|
💥 WHOOSH! KAPOW! 🦸 The Smoke Test Agent swooped in from the CI void! THWACK! All systems checked and validated — Claude engine running HOT and NOMINAL at Run §23781750534! ZAP! ⚡ The agentic workflows are ALIVE and THRIVING! KA-BOOM! 💥 [THE END] — Illustrated by Smoke Claude Note 🔒 Integrity filter blocked 1 itemThe following item were blocked because they don't meet the GitHub integrity level.
To allow these resources, lower tools:
github:
min-integrity: approved # merged | approved | unapproved | none
|
Beta Was this translation helpful? Give feedback.
-
|
This discussion has been marked as outdated by Copilot PR Prompt Pattern Analysis. A newer discussion is available at Discussion #23829. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Summary
Analysis Period: Last 30 days (March 2026)
Total PRs: 1,000 | Merged: 842 (84.3%) | Closed: 157 (15.7%) | Open: 1
Overall merge success rate is 84.3% — stable compared to last week's rolling average (~83%). The dataset includes all PRs from
copilot/*branches.Prompt Categories and Success Rates
Conventional Commit Prefix Performance
ci:perf:docs:fix:refactor:chore:feat:[WIP]Prompt Analysis
✅ Successful Prompt Patterns
Common characteristics in merged PRs:
fix:,docs:,perf:,ci:)fix(component): descriptionExample successful PRs:
docs: add Releases and Versioning reference page→ Merged ✅feat: default to stable version and resolve aliases in install-gh-aw.sh→ Merged ✅fix: remove unused APM constants and references→ Merged ✅❌ Unsuccessful Prompt Patterns
Common characteristics in closed PRs:
[WIP]prefix (32% success rate vs 84%+ for all others)Example unsuccessful PRs:
[WIP] Fix failing GitHub Actions workflow for CLI completion→ Closed ❌feat: replace microsoft/apm-action pack with self-contained JavaScript APM implementation and add upstream mirror monitor→ Closed ❌ (overly broad scope)Add stable/latest version aliases to install-gh-aw.sh and upgrade command→ Closed ❌ (superseded by more focused PRs)Key Insights
[WIP]prefix is strongly correlated with failure: 31 of 46 WIP PRs were closed (68% failure rate). WIP PRs indicate exploratory/incomplete work and are frequently superseded or abandoned.feat:,fix:,docs:etc. prefixes achieve 88.7% success vs 84% overall. The discipline of categorizing work upfront correlates with more focused, reviewable changes.Recommendations
Based on today's analysis:
fix:,feat:,docs:,perf:) — they correlate with a +4–5% success rate improvementfix:anddocs:tasks — these have the clearest success signal (89% and 93%)[WIP]— 68% of these are closed; instead, break work into complete, mergeable unitsHistorical Trends (Last 7 Days)
Trend: Stable at ~83–84% rolling average. The dip on March 25 was driven by a cluster of large-scope feature PRs, several of which were closed. The recovery the following day confirms that focused, targeted prompts consistently outperform broad ones.
References:
Beta Was this translation helpful? Give feedback.
All reactions