Peer Review

Query multiple LLM CLIs in parallel and collect their responses.

Available CLIs

All three should be invoked via the Bash tool as background tasks, run in parallel:

• Claude: unset CLAUDECODE && claude --model opus -p "$PROMPT"
• Codex: codex exec -m "gpt-5.5" --skip-git-repo-check "$PROMPT"
• Gemini: gemini -m "gemini-3.1-pro-preview" -p "$PROMPT"

Self-invocation rule

The orchestrating agent must NOT call its own CLI via Bash — it will fail or produce empty output. Use a subagent for your own model's contribution and Bash CLI commands for the other models.

Workflow