Health Report

Run a health check of the AI Harness system.

Live System State

!ps aux | grep -E "(bot\.ts|heartbeat-runner|claude)" | grep -v grep 2>/dev/null || echo "(no processes)" !launchctl list | grep com.aiharness 2>/dev/null || echo "(no launchd agents)" !ls -lh bridges/discord/harness.db 2>/dev/null || echo "(no database)" !tail -3 heartbeat-tasks/logs/*.log 2>/dev/null | head -20 || echo "(no logs)"

Check Modules

Parse $ARGUMENTS to determine scope. Default is full (all checks).

Bot Check (bot or full)

1. Check if bot process is running: look for bot.ts in process list (from live state above)
2. Check PID file: bridges/discord/.bot.pid — does it exist? Is the PID alive?
3. Check database exists and is readable: bridges/discord/harness.db
4. Report: running/stopped, PID, uptime estimate

Database Check (db or full)

1. Check harness.db exists and size
2. Run sqlite3 harness.db "SELECT name FROM sqlite_master WHERE type='table'" to verify schema
3. Check WAL file: harness.db-wal existence and size
4. Count rows in key tables: sessions, channel_configs, task_queue, dead_letter
5. Report: table counts, file size, WAL status

Heartbeat Check (heartbeat or full)

1. List LaunchAgent plists: ~/Library/LaunchAgents/com.aiharness.heartbeat.*.plist
2. For each, check if loaded in launchctl (from live state above)
3. Read state files: heartbeat-tasks/*.state.json — check last_run, consecutive_failures
4. Flag any tasks with consecutive_failures > 0
5. Report: task name, status (loaded/unloaded), last run, failures

Vault Check (vault or full)

1. Count files in vault/learnings/ by type prefix (LRN/ERR/FEAT)
2. Check for orphaned wikilinks: grep for [[ references and verify targets exist
3. Check for entries missing required frontmatter fields
4. Count entries by status
5. Report: file counts, any issues found

Truncation Check (truncation or full)

1. Read context-log/truncation-stats.json if it exists
2. Report: events by source, average % lost, critical count
3. Flag any sources with critical truncation events
4. Read last 10 lines of context-log/truncation-events.jsonl for recent events
5. Report: worst truncation sources, whether structure damage occurred

Quick Check (quick)

Run only: bot process alive? Database exists? Any heartbeat failures? One-line summary.

Output Format

=== AI Harness Health Report ===

Bot:       RUNNING (PID 12345)
Database:  OK (harness.db, 256KB, 5 tables)
Heartbeat: 2/2 tasks loaded, 0 failures
Vault:     42 learnings (38 resolved), 0 issues
Truncation: 15 events, 0 critical, avg 12% lost

Overall: HEALTHY

For issues, add a ⚠ Issues section listing each problem and suggested fix.