mirror of
https://github.com/kevinveenbirkenbach/computer-playbook.git
synced 2026-11-10 12:47:04 +00:00
kevinveenbirkenbachandClaude Opus 5.5
8458f8043a
Deploy jobs killed by the deploy cap left no record of host memory or CPU pressure, so a slow job could not be told apart from a starved one. Every deploy matrix job now starts a background sampler that appends mem, load, PSI and the ten largest processes by RSS (pid, rss, comm only; never argv or env) to host-metrics.log in the rescue-diagnostics folder every 60 s, which the existing always() upload ships. report_failures now downloads each failed artifact into its own folder, because gh refuses to overwrite and the second artifact of a role was lost, and it reports 'no rescue snapshot' only when no artifact holds capture content below <distro>/<app>. Accepted MEDUSA false positive: B603 on the list-argv gh subprocess call in cli/meta/ci/report_failures.py. 🎯 Fix: 90 % 💥 Break: 3 % 💸 Cost: – ⏱️ Time: – 🛡️ Security: 100 % 🏆 Score: 87 🚦 Gate: ✅ 90/5, cost and time <= 100, security 100 % Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>