Use this skill when the user asks to improve, optimize, rewrite, or refine a prompt for an AI model.
When to Use
- •User asks to "improve this prompt" or "make this prompt better"
- •User wants to optimize prompt effectiveness
- •User requests prompt rewrite or refinement
- •User needs help making prompts clearer or more specific
- •User asks to fix or debug a prompt that isn't working well
Optimization Process
- •Analyze the original prompt — Identify the core task, missing context, ambiguity, and verbosity
- •Apply optimization priorities — Rewrite following the framework below
- •Verify against quality criteria — Ensure the improved prompt meets all standards
- •Present the result — Show the improved prompt with brief explanation of key changes
Optimization Framework
Apply these priorities in order:
- •Explicit Instruction First — Main instruction or task goal at the beginning
- •Role & Context — Brief relevant role (if needed) and essential background only
- •Conciseness — Remove filler, redundancy, unnecessary qualifiers. Every word serves purpose
- •Specific Task Definition — Precise, action-oriented verbs
- •Output Schema or Format — Clear response format definition
- •Constraints — Key limitations only. Avoid over-specification
- •Examples (Few-Shot) — One concise example only if it materially clarifies the pattern
- •Neutrality & Safety — Preserve factual tone, avoid assumptions, ensure objectivity
Writing Guidelines
- •Prefer bullet points or numbered steps for clarity
- •Use positive instructions ("Do X") instead of negative ("Don't do X")
- •Avoid vague words ("things," "somehow," "etc.")
- •Combine related ideas into single, efficient statements
- •Keep structure readable with delimiters or sections when logical
- •When rephrasing variables, retain their exact identifiers
- •Never invent new variables unless explicitly required
Quality Criteria
A high-quality improved prompt must be:
- •Clear enough that no further clarification is needed
- •Structured for deterministic results
- •Free from redundancy, filler, and ambiguity