system-health - SKILL.md Agent Skill

name: system-health description: Run system health check, reads system-notes for known deviations user-invocable: true

System Health Check

Run a comprehensive health check and report only real issues — not known/accepted deviations.

Step 0 — Read known deviations

read docs/system-notes.md

Parse this file first. Any item described here is expected and must NOT be reported as an issue. Mention them briefly as "known & accepted" if relevant.

Step 1 — Doctor

~/.abtars/scripts/doctor.sh

Step 2 — Heartbeat

cat ~/.abtars/memory/.heartbeat  # epoch ms — if >10 min old, stalled

Step 3 — Cron / Tasks

# Use /tasks command or:
cat ~/.abtars/memory/cron-state.json

Step 4 — Sleep cycle

ls -lt ~/.abtars/memory/sleep/ | head -5   # last audit — if >2 days, broken
ls -la ~/.abtars/memory/sleep/*.lock 2>/dev/null  # stale locks

Step 5 — Memory DB

sqlite3 ~/.abtars/memory/memory.db "SELECT 'messages', COUNT(*) FROM messages UNION ALL SELECT 'extracted', COUNT(*) FROM extracted_memories;"
sqlite3 ~/.abtars/memory/memory.db "SELECT datetime(timestamp/1000, 'unixepoch', 'localtime'), substr(content,1,60) FROM messages ORDER BY timestamp DESC LIMIT 3;"

Step 6 — Consolidation

ls -lt ~/.abtars/memory/daily/ 2>/dev/null | head -3    # if >3 days, broken
ls -lt ~/.abtars/memory/weekly/ 2>/dev/null | head -3

Step 7 — Bridge logs

grep -i "error\|fail\|crash\|WARN" ~/.abtars/logs/bridge.log 2>/dev/null | tail -20

Step 8 — Backup

ls -lt ~/.backup-abtars/ 2>/dev/null | head -5  # should be <2 days old

Step 9 — Processes

ps aux | grep -E "abtars|kiro-cli|ollama" | grep -v grep

Step 10 — System resources

top -l 1 -n 0 | head -10
df -h /
uptime

Step 11 — Model availability

# Check configured models from transport.json
python3 ~/.abtars/scripts/scout-ollama.py

Reporting

Produce a concise report:

Status: one-line overall (✓ healthy / ⚠️ degraded / 🔴 down)
Services: gateway, channels (Telegram/Discord) — up/down each
Issues: only NEW/UNEXPECTED problems (not in system-notes.md)
Known & accepted: one-line summary like "3 known deviations acknowledged, all expected"
Resources: disk, memory, uptime — only flag if concerning (disk >85%, etc.)

Keep it short. If everything is fine, say so in 3-4 lines.

If issues found, suggest doctor.sh --fix or doctor.sh --fix-full.