AI prompts for ai agent framework

Require verification-first behavior to reduce unsupported or hallucinated answers.

This prompt checks a prompt written for an LLM for unclear wording, missing definitions, instructions that clash, and assumptions left unstated. It returns a list of issues, suggested edits to specific lines, and a rewritten version.

Single pasteable recipe that drafts, critiques, then revises against criteria.

Classify items into labeled categories using few-shot examples and confidence scores.

This prompt rates an AI answer on how well it follows its original instructions, measured against what a top human expert would deliver. It returns a letter grade from A+ to F, five reasons for the grade, a score out of 100, and how the score was reached.

Grade an LLM or human output against a custom rubric.

Rewrite a weak prompt into a clearer, higher-signal prompt.

Produce multiple JSON objects inside named XML tags for structured multi-object extraction.

Add a system-prompt block defining when an agent should run independent tool calls in parallel.

This prompt designs an AI agent to take over a manual business process you describe. It starts with clarifying questions, then works through 15 steps covering stages, tools, memory, the main loop, checks, error handling, human approval points, monitoring, tests, and a build plan.

This prompt rewrites a loose LLM prompt into a Markdown system prompt with set sections: identity and purpose, steps, output instructions, examples, and an input slot. It follows the layout used by Fabric patterns.

Tell an agent to prioritize user instructions over skills and tools.

Structure a complex LLM task with XML tags for context, constraints, and output format.