Everything durable → cloud

Every long-lived artefact (knowledge, skills, memory, secrets, repo mirrors) has a cloud copy; local machine is a cache, not source of truth.

Everything durable → cloud

Rule

Every durable artefact has a cloud copy. Local machine is a cache, not source of truth. Reproducibility requires cloud-primary posture.

What counts as durable

What does NOT need cloud

• Runtime state (open files, running processes, TUI history)
• Machine-specific paths (C:\D\oriz\...)
• Cached indexes (regenerate on demand)
• .mimo/, .opencode/, etc. — dropped agent state, dropped rule

Why cloud-primary

• Machine loss — corp laptop wipe or personal laptop failure = zero-loss recovery.
• Cross-machine parity — every laptop can be primary in <5 min bootstrap.
• Zero-fleet failover — single-agent (CC only) means no in-fleet backup; cloud IS the backup.
• Public discovery — knowledge/ + skills/ have value beyond us; publish them.

Constraints

• No card-on-file — cloud targets must be free-tier per no-card-on-file.
• sops+age encrypted — anything sensitive travels encrypted.
• Git as sync target when possible — vendor-neutral, portable.

Anti-patterns

• ❌ Local-only artefact that can’t be recreated from cloud
• ❌ Cloud target that requires card
• ❌ Sync without encryption for sensitive data
• ❌ “I’ll push it later” — auto-sync hooks per auto-sync-enabled-2026-07-03

Cross-refs

• cloud-publish-knowledge-2026-07-03
• memory-cross-machine-sync-2026-07-03
• no-card-on-file

Edit on GitHub · Back to index