69. Prompt Enhancing Techniques

The lesson

The core idea: before you use an important prompt in a real chat session, you first have a conversation WITH the AI ABOUT the prompt. The AI acts as an expert prompt engineer, reviewing and improving your draft. Only the final enhanced prompt is used for real work - in a brand-new session.

Why this works: an LLM is excellent at spotting what is missing from a prompt - ambiguous instructions, missing context, no output format, an undefined audience. You get that review for free by simply asking for it. It is like having a circuit design reviewed by a colleague before you etch the board: the review conversation makes the final design better.

The technique has four parts, exactly as implemented in TPEE's 5-step Prompt Enhancing Workflow strip:

Part 1 - The Pre-Prompt. A short text placed BEFORE your prompt. It tells the AI: 'I am preparing a prompt that I will use later in a fresh chat session. Act as an expert prompt engineer: analyse my prompt, ask me questions, and help me improve it. Do NOT execute or answer the prompt itself.' Without this, the AI would simply try to answer your draft. The Pre-Prompt ends with two marker lines: '-----End Of Pre Prompt-----' and '-----Start of Prompt to Enhance-----'.

Part 2 - The Post-Prompt. A short text placed AFTER your prompt. It starts with the marker lines '-----End of Prompt to Enhance-----' and '-----Start Of Post-Prompt-----', then tells the AI how to run the enhancement conversation and where to save the result: 'When I tell you to SAVE, save the final enhanced prompt as a plain text file to the directory and file name I give you.' The marker lines work like start/end delimiters in a serial protocol frame - the AI always knows exactly where your prompt begins and ends.

Part 3 - The enhancement conversation. Combine Pre-Prompt + your prompt + Post-Prompt into one block (TPEE's COMBINE button does this to the clipboard in one click) and paste it into the AI chat. The AI will ask clarifying questions: What is the audience? Which output format? What constraints? Answer them honestly - every answer becomes specific text in the enhanced prompt. When you are satisfied, tell the AI to SAVE the enhanced prompt to your file (for example MyProject_Enhanced.txt next to your prompt file).

Part 4 - Verify and reuse in a FRESH session. Open the saved enhanced prompt, read it, and fix anything the AI misunderstood (TPEE's Enhanced Prompt Editor at step 5 does this). Then copy it and paste it into a brand-new chat session. The fresh session matters: the enhancement chat is full of drafts and meta-discussion that would pollute the real run. A clean context means the model sees ONLY the polished prompt.

Expected results: prompts become noticeably more specific - vague verbs are replaced with measurable instructions, missing context is filled in, output format and audience are stated explicitly. First-answer quality goes up and you need fewer retry iterations, which saves tokens and money on every future reuse of that prompt. The enhanced prompt file also becomes a reusable library asset.

When to use it: any prompt you will reuse many times, any high-stakes prompt (code generation for production, business documents, PRDs), or any prompt that keeps producing mediocre answers. For quick one-off questions the overhead is not worth it - just ask directly.

In TPEE: tick USE PROMPT ENHANCING WORKFLOW in the top panel and follow steps 2 to 5 - edit Pre-Prompt, edit Post-Prompt, COMBINE to clipboard, then open/verify/copy the enhanced prompt. Pre/Post-Prompts are saved next to your prompt file (_Pre_Pmpt.txt / _Post_Pmpt.txt) and auto-load next time. See User's Manual chapter 28 for full details.

Check yourself