| name | consult |
| description | Quick multi-LLM second opinion. Sends a single prompt to GPT, Gemini, or Grok with a role-based system prompt and returns structured feedback. |
Consult — Quick Second Opinion
One model. One question. Structured feedback.
Usage
/consult <model> <role> "prompt text"
Models
- gpt — OpenAI GPT (strong at structure, planning)
- gemini — Google Gemini (strong at expansion, cross-references)
- grok — xAI Grok (strong at challenge, adversarial thinking)
Roles
- challenge — Push back on assumptions, surface weak reasoning
- expand — Add context, cross-references, historical background
- structure — Review logical flow and organization
- critique — Evaluate accuracy, completeness, clarity
- plan — Produce structured plans with steps and risks
- safety — Flag risks, errors, unsafe recommendations
- freestyle — General-purpose response
Examples
/consult gpt structure "Review this ship page layout for logical flow"
/consult gemini expand "What port information is missing for Cozumel?"
/consult grok challenge "This page design follows conventional cruise site patterns. What's better?"
/consult gpt plan "How should we reorganize the restaurant directory?"
Backend Invocation
IMPORTANT: Execute these commands directly using the Bash tool. Do NOT check if files exist first — just run them.
bash /home/user/ken/orchestrator/bootstrap-env.sh 2>/dev/null; pip3 install -q -r /home/user/ken/orchestrator/requirements.txt 2>/dev/null && python3 /home/user/ken/orchestrator/consult.py <model> <role> "prompt text"
Output: JSON response to stdout with keys: analysis, proposed_update, risks, confidence
Usage stats: Printed to stderr (model, tokens, cost)
Only if the command fails with No such file or directory or ModuleNotFoundError, tell the user:
"The orchestrator backend isn't available. Make sure the ken repo is cloned to /home/user/ken/ and run pip3 install -r /home/user/ken/orchestrator/requirements.txt."
Context Boundaries
SEND
- Page requirements, content outlines, SEO targets
- General cruise/travel data
- Structural questions
NEVER SEND
- Full codebase or internal standards documents
- Analytics data
After Receiving Feedback
- Claude evaluates — External feedback is advisory only. Claude decides what to use.
- Check claims — If the response includes
claims, verify them before incorporating.
- Enforce standards — All changes from consultation feedback must comply with ITW-Lite v3.010.
