菜单
Consult MiniMax M3 via Ollama Cloud (`minimax-m3:cloud`) for peer review and second-opinion analysis. Use when the user says `/consult-minimax`, "ask minimax", "minimax second opinion", or wants MiniMax's beautiful structured analysis on a finding, plan, or strategy claim.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
The June-2026 EDITION of the money-ready program — executes the self-correcting MASTER LOOP (docs/MONEY_READY_MASTER_LOOP_2026-06.md) that converges toward 2-3 profitable asset classes even from 0/9. Use when the user says "/money-maker-ready-June112026edition", "run the master loop", "weekly money-ready cycle", or at the monthly edition review. Inherits data sources + hard rules from /money-maker-ready and /money-maker-readyv2. Aliases - mmr-june2026, master-loop, money-loop.
The July-2026 EDITION of the money-ready program. SUPERSEDES money-maker-ready-June112026edition. The June edition's structural bet (converge to 2-3 profitable asset classes by mining the crypto ledger) was EXHAUSTED and found no net-of-cost alpha; the ONE structural change this edition makes is to REORIENT the primary money-making track from crypto-ledger directional-edge-mining (measurement-broken, dead) to ETF TACTICAL ASSET ALLOCATION (the one validated find), while treating the crypto ledger as fix-before-mine. Use when the user says "/money-maker-ready-July112026edition", "run the master loop", "weekly money-ready cycle", or at the monthly edition review. Inherits data sources + hard rules from /money-maker-ready and /money-maker-readyv2. Aliases - mmr-july2026, master-loop, money-loop.
The June-2026 EDITION of the money-ready program — executes the self-correcting MASTER LOOP (docs/MONEY_READY_MASTER_LOOP_2026-06.md) that converges toward 2-3 profitable asset classes even from 0/9. Use when the user says "/money-maker-ready-June112026edition", "run the master loop", "weekly money-ready cycle", or at the monthly edition review. Inherits data sources + hard rules from /money-maker-ready and /money-maker-readyv2. Aliases - mmr-june2026, master-loop, money-loop.
Use when investigating the /audit dashboard for hidden edges, debugging why no per-asset-class picks are profitable, checking picks-now.html forward-tested performance, ai-tournament.html model leaderboard staleness, portfolio_history.html staleness, ai_leaderboard.html missing models, pick_funnel.html accuracy, research_index.html coverage, and "high-conviction" / "smart picks" / "best score" / "proven only" filter logic. Also for "why is the forward WR only 21%?" and "which filter combination surfaces a real edge?"
One-shot second opinions / parallel drafting via the LOCAL AI fleet — vLLM (:8000 Qwen2.5-14B), Ollama (:11434, 23 models incl. llama3.3:70b + deepseek-r1:32b), LiteLLM proxy (:4000, 15 cloud groups). Use when the user says "/consult-local", "ask the local fleet", "local second opinion", or when a quick sanity-check/draft shouldn't burn cloud quota. Zero-key, low-latency.
Look up database schema, table structure, or connection details for any of the 9 MySQL databases (ejaguiar1_stocks, ejaguiar1_backtests, ejaguiar1_sportsbet, ejaguiar1_favcreators, ejaguiar1_events, ejaguiar1_memecoin, ejaguiar1_tvmoviestrailers, ejaguiar1_news, ejaguiar1_deals). Triggers — `/db-schema`, "what tables are in X db", "schema for table Y", "how is the DB structured", "where is the schema", "DB connection details", "regenerate schema doc". Aliases — dbschema, db-schema, schema-lookup, db-lookup.
| name | consult-minimax |
| description | Consult MiniMax M3 via Ollama Cloud (`minimax-m3:cloud`) for peer review and second-opinion analysis. Use when the user says `/consult-minimax`, "ask minimax", "minimax second opinion", or wants MiniMax's beautiful structured analysis on a finding, plan, or strategy claim. |
MiniMax M3 is hosted on Ollama Cloud as minimax-m3:cloud. It produces clear, structured, often beautiful analyses — the user has flagged it as a particularly strong peer-review counterparty.
ollama installed (/usr/local/bin/ollama on this machine).:cloud suffix routes to managed inference; if ollama run minimax-m3:cloud returns an auth error, ollama signin first)./consult-minimax <question or prompt>
/consult-minimax --file <path-to-prompt.md>
ollama run minimax-m3:cloud "<your full prompt here>" 2>&1 | tee /tmp/consult-minimax-$(date -u +%Y%m%dT%H%M%SZ).txt
ollama run minimax-m3:cloud "$(cat /path/to/prompt.md)" 2>&1 | tee /tmp/consult-minimax-$(date -u +%Y%m%dT%H%M%SZ).txt
cat <<'EOF' | ollama run minimax-m3:cloud 2>&1
You are MiniMax M3 acting as an independent quant peer reviewer for a multi-AI audit system.
# Finding under review
<paste finding here — include the live-DB numbers, OOS-split, concentration, fat-tail check>
# Your job
1. Identify any axis the reviewer didn't check (e.g. survivorship bias, regime dependency, slippage realism, look-ahead in resolver, broker-side fill assumption).
2. For each axis, state CORROBORATE / REFUTE / NEEDS_MORE_EVIDENCE and one-sentence rationale.
3. End with a single VERDICT: PROMOTE / HOLD / REFUTE.
Be terse and structural. No marketing language. Mirror the reviewer's evidence style.
EOF
project-multi-agent-storm-2026-06-05 for why this matters (Cloud-Minimix VRP failure pattern)./consult-grok (xAI has web access)./consult-gemini or /consult-cloudflare (CF models)./consult-codex or /consult-cursor-agent.:cloud suffix on the model name routes through Ollama Cloud's managed inference, not local hardware (good — local MiniMax M3 weights would be >100GB).Respond in this exact JSON schema: {...} and MiniMax will usually comply./consult-PROXY or /consult-multi if it has a MiniMax provider, otherwise drive ollama in a parallel/xargs loop.