P2: Define safe unification plan for .agent/.opencode/.claude assets

[emil14]

Context

Repository agent-related assets are currently split across multiple roots:

• .agent (local tasks/notes/skills/workflows)
• .opencode (skills/agents used by CI and OpenCode)
• .claude (authoring rules)

This works today, but introduces discoverability and governance overhead.

Goal

Design a safe unification strategy for agent assets without breaking current CI/CD behavior.

Non-goals

• No immediate migration in this task.
• No sync scripts/mechanisms yet.

Deliverables

1. Inventory of asset types and current consumers (local Codex, OpenCode CI, human docs).
2. Target directory model (single source of truth + compatibility constraints).
3. Migration plan with rollback steps.
4. Minimal validation checklist for CI safety.

Acceptance criteria

• Proposed model keeps existing workflows operational during transition.
• Explicit compatibility story for OpenCode agents/skills in .opencode.
• Clear phased rollout with ownership.