メニュー
Run system health checks, auto-repair issues, and generate health dashboard. Use for /patrol, /health, or when diagnosing system problems.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when generating any HTML dashboard or view for Software of You. Provides the design system, base template, and component patterns for consistent visual output across all modules.
Use when checking which modules are installed, resolving cross-module features, or handling module-aware queries. This skill explains how modules discover each other and activate enhanced features.
Use when performing any database read or write operation in Software of You. This skill provides the schema reference, computed-view guidance, query patterns, and conventions for interacting with the SQLite database.
Use when processing meeting transcripts, analyzing conversations, extracting commitments, or generating communication coaching insights. This skill provides parsing patterns, coaching guidelines, and the relationship scoring algorithm.
Use when generating project pages, project briefs, or assessing project health. This skill provides the scoring methodology for momentum, risk, action prioritization, and client relationship temperature.
| name | health-monitor |
| description | Run system health checks, auto-repair issues, and generate health dashboard. Use for /patrol, /health, or when diagnosing system problems. |
Self-healing health checks for Software of You. Validates infrastructure, auto-repairs what it can, and generates a visual health dashboard.
/patrol — run all checks, auto-fix issues, generate health dashboard/health — quick status check (read-only, no repairs)Run the health check script:
# Full patrol with auto-repair + dashboard
python3 "${CLAUDE_PLUGIN_ROOT:-$(pwd)}/scripts/health_check.py" --fix --dashboard
# Read-only check
python3 "${CLAUDE_PLUGIN_ROOT:-$(pwd)}/scripts/health_check.py"
# JSON output for programmatic use
python3 "${CLAUDE_PLUGIN_ROOT:-$(pwd)}/scripts/health_check.py" --fix --json
| Check | What | Auto-Fix |
|---|---|---|
| Database integrity | 33 required tables + 8 required views exist, SQLite integrity_check passes, no FK violations | Runs bootstrap to recreate missing tables/views |
| OAuth tokens | Token files exist, not expired, have refresh tokens | Auto-refreshes expired tokens |
| HTML views | Core views (dashboard, contacts, week-view, email-hub, nudges, timeline) exist with content | Flags missing views in soy_meta for regeneration |
| Backups | Backup directory has files, latest is < 48h old | Creates fresh backup |
| Sync freshness | Gmail, Calendar, Transcripts synced within 60m | Reports staleness (sync via pipeline) |
If the patrol flags HTML views for regeneration, regenerate them:
SELECT value FROM soy_meta WHERE key = 'views_need_regeneration';
If this returns a list, generate those views using the dashboard-generation skill.
Before any destructive database operation (wipe, reset, drop), always call:
from scripts.health_check import enforce_backup_before_destructive
enforce_backup_before_destructive("database_wipe")
Or via bash:
python3 -c "from scripts.health_check import enforce_backup_before_destructive; enforce_backup_before_destructive('your_operation')"
The patrol generates output/health.html showing: