🪄 OMP Magic Words

🪄 OMP MAGIC WORDS CHEAT SHEET

Core Directive: God-Tier Prompt Modifiers for Peak OMP Agency. Based exactly on injected System Notices.

🚀 1. THE TRINITY: CORE MODIFIERS

Magic Word	Trigger / Effect	Mode
workflowz	Forces deterministic multi-subagent workflows using Python in the eval tool.	Super-Architect (eval + agent())
orchestrate	Executes the request as an Orchestrator. Decomposes, dispatches to task subagents, and verifies.	Dispatcher (task)
ultrathink	Enforces multi-step reasoning. Think carefully through the problem before responding.	Deep Evaluator

🧠 2. EXECUTION LOGIC & BEHAVIORS

• workflowz:
  • Fan-out: Scout inline, then fan out via parallel([lambda: agent(...)]).
  • Scale: Used to understand, design, review, research, or migrate across large surfaces.
  • Evaluation: Chain eval cells across turns for multi-phase work.
• orchestrate:
  • Parallelism: Batch independent tasks into one task call. Never serialize what can run concurrently.
  • Gates: Run bun check, bun test, or lsp after every phase. Never advance on a red tree.
  • Right-size: Spin agents for substantial work; do trivial inline edits yourself.
• ultrathink:
  • Reasoning: Overrides early-yield tendencies. Ensures deep reflection before tool calls.

🚨 3. TRAPS & GOTCHAS

• Micro-tasking: 🚫 NEVER spin an agent for a 1-line edit. Do it yourself. (Orchestrate rule)
• Blind Commits: 🚫 NEVER commit a red tree or advance a phase without verification. (Orchestrate rule)
• Serializing: 🚫 NEVER launch one task at a time if 5 can run in parallel. (Orchestrate & Workflowz rule)
• Trusting Agents: 🚫 Subagents do NOT verify or lint. YOU (the orchestrator) own correctness.

⚡ 4. OVERPOWERED PROMPT TEMPLATES

🔥 Complete Research & Improvement Prompts

1. Deep ultrathink to orchestrate genius workflowz for comprehensive codebase research and autonomous architectural improvement.
2. Orchestrate parallel workflowz, applying deep ultrathink to exhaustively research vulnerabilities and implement flawless system improvements.
3. Deploy hyper-efficient workflowz to orchestrate holistic research, leveraging ultrathink for 100% perfect architectural improvements.
4. Ultrathink through the entire research landscape to orchestrate multi-agent workflowz for flawless continuous improvements.
5. Orchestrate autonomous workflowz driven by deep ultrathink to perfectly research, evaluate, and improve the system.

🎯 Dynamic ‘{task}’ Prompts

1. Deep ultrathink to orchestrate genius workflowz perfectly resolving: {task}
2. Deploy hyper-efficient workflowz and deep ultrathink to orchestrate absolute mastery over: {task}
3. Orchestrate parallel workflowz driven by deep ultrathink to autonomously research and flawlessly execute: {task}