diag-infra

name	diag-infra
description	Infrastructure diagnosis with differential-first approach. Gathers evidence, ranks hypotheses, proposes cheapest test.
user-invocable	true
allowed-tools	Bash, Read
truth_contract	{"canonical_sources":["ops/state/config/repo-map.json"],"live_load_required":["kubectl get nodes -o wide 2>/dev/null || true","ssh ubuntu@<host> 'systemctl status <service>' 2>/dev/null || true"],"examples_only":[],"never_hardcode":["cluster IPs or host names (read from repo-map.json)","assumed service state (always verify live)"]}

Diagnose infrastructure issues using a differential-first approach. Gather evidence FIRST, then hypothesize.

Steps

1. Gather evidence (no fixes yet):
   • dmesg --time-format iso | tail -120
   • journalctl -p err --since "24 hours ago" | tail -200
   • free -h
   • df -h
   • top -bn1 | head -40
   • ss -tlnp | head -20
2. Produce ranked differential:
   • 3 hypotheses with evidence FOR, evidence AGAINST, and cheapest test for each
3. Propose 1 cheapest test to run next.

Important

• Do NOT modify ICN repo code during infra diagnosis.
• Do NOT start fixing things before the differential is complete.
• If the issue appears to be ICN application-level (not infra), say so and recommend switching to a code debugging session.