CLI Bridges

CLI bridges are lightweight commands that delegate questions to a different AI CLI or model outside the current OMX session. The current direct bridge skills are $ask-claude and $ask-gemini; broader multi-model work should be handled by explicit team or workflow orchestration rather than an undocumented bridge name.

What it is

Each bridge spins up the target CLI binary as a fresh subprocess, passes along your question or task, and feeds the answer back into the current OMX session. The subprocess starts clean. With none of your context inherited, the call stays walled off from whatever you're working on, which is exactly the point: an outside perspective that doesn't contaminate the session you're in.

Bridge patterns

• $ask-claude — Sends the question directly to a Claude CLI subprocess. Use it for quick lookups, summarization, or fetching Claude's reasoning without touching the Codex context.
• $ask-gemini — Launches a Gemini CLI subprocess. Especially useful for large-context lookups or when you want to compare with Google's model.
• Team-level mixed CLI workers — For broader multi-model coverage, configure team workers deliberately and keep the synthesis in the main OMX workflow.

Example

Single-model consultation:

codex
> $ask-claude "what is the recommended retry strategy for idempotent HTTP endpoints?"

codex
> $ask-gemini "summarize the tradeoffs between RSA and ECDSA for JWT signing"

Coordinated multi-model review:

codex
> $team 2:executor "compare Claude/Gemini-informed options for a public REST API rate limiter"

Calling directly from a shell script without going through the skill layer:

# Ask Claude directly from a shell script
echo "explain this error: $(cat error.log)" | claude --print

Related

• ask-claude — $ask-claude skill reference
• ask-gemini — $ask-gemini skill reference
• MCP — Model Context Protocol for structured tool invocation
• OpenClaw — event bridge for cross-session notifications