This is a heavyweight, explicit-opt-in workflow. Apply this gate before opening any
reference, creating a review log, or contacting another agent.
Activate only when the current user request directly asks to run this consultation, or
uses the skill's canonical host invocation:
in Codex,
/dev:consult-outside-expert
from this Claude Code plugin, or
when installed as a standalone Claude Code skill. An
imperative request to ask an independent outside expert or agent for a second opinion
also qualifies.
An already-running consultation is exempt. If
or
exists for the current consultation, it was already explicitly started, so continue it
when the user supplies a requested mediator decision or asks to resume. A reply such as
"Option A" after a decision-point prompt applies that decision and continues the existing
session; the explicit-opt-in requirement governs initial activation, not continuation.
If the gate fails, answer normally and stop. Do not open the guide, spawn an expert, or
create consultation artifacts.
Open
and follow it. Do not proceed without it.
When the activation gate passes, consult an outside expert to collaboratively refine
work through iterative back-and-forth. It supports: