Second-model AI auto-review for DeepSeek Harness approval requests: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default, fully auditable from the session log (approval/asked -> autoReview/verdict -> approval/decided).

The ultimate 2-way bridge between OpenAI Codex and DSH: bidirectional chat sync, live Skills mount & auto MCP mirroring.