Lightrun AI

A collection of reusable skills for investigating runtime issues with Lightrun.

What is in this repository

• skills/ — skill folders (SKILL.md, plus optional assets/scripts)
• .cursor-plugin/ — Cursor plugin metadata
• .claude-plugin/ — Claude plugin metadata
• .cursor/mcp.json — Cursor MCP server configuration used by the plugin metadata

Available skills

• lightrun-live-runtime-debugging — deterministic live-runtime investigation workflow with preflight checks, evidence capture, and handoff output.
• lightrun-error-remediation-automation — error remediation automation workflow with root cause analysis, mitigation strategies, and automated fixes.
• lightrun-ask-prod — live production runtime question answering for current values, durations, hit counts, and distributions.
• lightrun-slow-execution-diagnosis — diagnosis for slow endpoints, latency or SLO regressions, timeouts, deadlines, and hangs.
• lightrun-runtime-aware-pr-review — pull request review workflow that combines live runtime samples with diff-aware risk analysis.

Demos

• Codex live debugging demo
• Claude live debugging demo

Installation

Cursor

1. Open Cursor.
2. Add plugin from repository URL (or local path).
3. Use this repository URL: https://github.com/lightrun-platform/lightrun-ai

Claude

This repository includes:

• .claude-plugin/plugin.json
• .claude-plugin/marketplace.json

Use your Claude plugin/marketplace flow to install from this repository.

Codex

Codex loads skills from .agents/skills locations.

1. Copy or symlink skill folders from skills/ into one of:
   • $REPO_ROOT/.agents/skills
   • $HOME/.agents/skills
2. Restart/reload Codex if needed.

Requirements

• Access to Lightrun MCP server
• Completed MCP OAuth authorization for your environment

Contributing

1. Add a new folder under skills/<skill-name>/.
2. Add SKILL.md with YAML frontmatter (name, description).
3. Add optional assets/scripts if needed.
4. Update marketplace/plugin metadata when adding publishable skills.