Prompt Patterns That Actually Work — Research, Meta-Prompting, and System Design You Can Use Today

The problem: Most prompting guides teach you to prompt a chat interface, not an agent. System prompt design for autonomous agents, meta-prompting for self-improvement, and research-grade prompt patterns are a different discipline — and they're not documented anywhere clearly.

The solution: Four prompting frameworks from production agent work — including the meta-prompt pattern for self-improving agents, the research mega-prompt for deep investigation tasks, and a cost calculator for making prompting decisions economically.

What you get (4 SKILL.md frameworks):

• Meta-Prompt — the prompt pattern that helps an agent improve its own system prompts
• Research Mega-Prompt — the prompt architecture for comprehensive deep-research tasks
• System Prompt Patterns — structural patterns for agent system prompts that actually hold up in production
• Agent Cost Calculator — the framework for estimating and optimising per-task inference cost

Who it's for: Operators building production agents who want the prompting patterns that distinguish reliable systems from fragile ones.