skills.rest
mattbutlerengineering
mattbutlerengineeringCOMMUNITY
Explore all skills by this author →
GitHub

learning-loop

Detect software regressions and create labeled GitHub issues from sensor metrics.

1|Updated Jan 24, 2026
Search Tags:#continuous improvement#regression detection#GitHub issues#threshold tuning#CI health#Sentry triage#session log analysis
One-click install
npx skills add https://github.com/mattbutlerengineering/mattbutlerengineering --skill learning-loop-mattbutlerengineering
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: learning-loop
Source: https://github.com/mattbutlerengineering/mattbutlerengineering/tree/main/.claude/skills/learning-loop
Command: npx skills add https://github.com/mattbutlerengineering/mattbutlerengineering --skill learning-loop-mattbutlerengineering

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires gh, node, fetch, child_process, fs, path, and includes scripts (resource) components.

What problem does it solve?

It reduces the time between detecting a regression and confirming what fixed it by running a sensor-driven closed-loop process that creates actionable issues and verifies outcomes.

Core Features & Use Cases

  • Sensor-driven regression detection: Runs a unified sensor report that gathers metrics from CI health, ACMM state, PR activity, Lighthouse inventory, Sentry errors, GitHub issue signals, and agent cost logs, then flags regressions for follow-up.
  • Fix verification for recent issues: Checks recently closed issues (labeled by originating sensor) and logs verification results, reopening when fixes fail to improve metrics.
  • Weekly skill-coverage discovery: On the configured day (default Friday), analyzes session logs to propose new skills when repeated patterns aren’t covered by existing skills.
  • Threshold self-tuning and loop logging: Computes false positive and fix effectiveness signals from recent verifications, writes improvement notes, and appends a dated summary log.

Quick Start

Invoke the learning loop by requesting: use /learning-loop to run sensor collection, create up to three regression issues, and verify recently closed fixes.

Frequently Asked Questions about learning-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automatically detect software regressions and create GitHub issues for them?▼

Automatically detect software regressions by running a sensor-driven Node script that collects CI health, Sentry errors, and GitHub issue signals, comparing them against baselines to create labeled regression issues with defined per-run caps.

What is closed-loop regression detection and how does it verify recent fixes?▼

Closed-loop regression detection verifies recent fixes by checking recently closed GitHub issues against current sensor metrics, logging verification results, and automatically reopening issues when fixes fail to improve the measured outcomes.

How do I triage Sentry errors and CI health metrics using a continuous improvement process?▼

Triage Sentry errors and CI health metrics by running a unified sensor report that gathers multiple system signals, flags identified regressions for follow-up, and persists the output to a JSON file for deduplication and remediation.

Can I use Node scripts to analyze session logs and propose new skills based on repeated patterns?▼

You can use Node scripts to analyze session logs on a configured weekly schedule, identifying repeated patterns not covered by existing skills to automatically propose new skill-coverage issues for continuous improvement.

Does the learning-loop threshold tuning automatically adjust regression detection sensitivity?▼

Threshold self-tuning automatically adjusts regression detection sensitivity by computing false positive and fix effectiveness signals from recent verifications, writing improvement notes, and appending a dated summary log.

What are the limitations of using GitHub searches to deduplicate regression issues?▼

Deduplicating regression issues with GitHub searches limits detection to existing repository data and requires predefined per-run issue creation caps, restricting the volume of actionable regression issues generated per execution.

Related Skills

Explore More Product & Management Skills→
dreamineeringdreamineering
7

depin-infrastructure-fetcher

Fetches real-time DePIN project data via CLI with multi-format outputs.

Community
Advanced
Atrium-HermesAtrium-Hermes

Self Review

Audit weekly agent outputs, reliability, and memory hygiene into a markdown report.

Community
Intermediate
luissimon96luissimon96

graph-bug-hunter

Automate bug discovery via static analysis, LSP diagnostics, and error catalog mining.

Community
Advanced
gagan114662gagan114662

github

Manage GitHub repositories, pull requests, issues, and Actions workflows via gh CLI and APIs.

Community
Intermediate