Alerts

Alerts

Overview

This skill provides comprehensive system health reporting by aggregating alerts from multiple sources. It surfaces issues that would otherwise be hidden in collapsed hook output or scattered across multiple files.

Usage

/​alerts           # Limited mode (default) - quick health check
/​alerts --full    # Full mode - comprehensive reporting

Modes

Limited Mode (Default)

Quick health check covering the most critical categories:

1. Code Health - Test failures, TypeScript errors, ESLint, pattern violations
2. Security - npm audit vulnerabilities, security patterns, secrets status
3. Session Context - MCP memory status, cross-session warnings, pending TODOs

Full Mode (--full)

Comprehensive reporting including everything in Limited plus:

1. Current Alerts - Deferred PR items, Backlog S0/​S1/​S2, encrypted secrets
2. Documentation Health - Orphaned docs, stale docs, CANON violations
3. Roadmap/​Planning - Overdue items, blocked tasks, stale SESSION_CONTEXT

Workflow

To run alerts check:

1. Determine mode (limited or full based on user request)
2. Run the appropriate checks using the script
3. Present results in a clear, actionable format
4. Offer to help resolve any issues found

Alert Categories

1. Code Health

2. Security

3. Session Context

4. Current Alerts (Full Mode)

5. Documentation Health (Full Mode)

6. Roadmap/​Planning (Full Mode)

Output Format

Present alerts grouped by severity:

## System Health Report

### Errors (must fix)
- [Code] 3 TypeScript errors in src/​components/​...
- [Security] 2 high vulnerabilities found

### Warnings (should address)
- [Code] 5 ESLint warnings
- [Alerts] 4 deferred PR items (2 aging)

### Info (for awareness)
- [Session] MCP memory is empty
- [Docs] 12 orphaned documentation files

Scripts

scripts/​run-alerts.js

Main script that runs all checks and outputs JSON results. Usage:

node .claude/​skills/​alerts/​scripts/​run-alerts.js --limited
node .claude/​skills/​alerts/​scripts/​run-alerts.js --full

The script outputs JSON that can be parsed and presented to the user.

Integration

This skill integrates with the Session Start Protocol in claude.md. At session start, Claude should:

1. Run /​alerts (limited mode) automatically
2. Present any errors or warnings to the user
3. Offer to run /​alerts --full for comprehensive check