Human-Agent Collaborative Scaffolding
What it does
A method of gradually creating complex structures through dialogue with an LLM agent. Instead of generating the entire result at once, the agent guides the user through stages: requirements gathering, structure development, and data population. At each step, the user controls the process and can make adjustments, which improves quality and reduces cognitive load.
When it helps
Use TableTalk when you need to create complex structured information (tables, databases, reports) in dialogue with an LLM. Particularly effective for tasks where requirements are not fully defined in advance or where control at each stage is important. Suitable for situations where direct generation produces unsatisfactory results due to task complexity. Ideal for data analysts, project managers, and researchers who need to structure information with the ability to make corrections at each step.
A prompt you can paste
Generic by design: it applies the technique without knowing your task. Adapt the marked parts.
You are a collaborative assistant tasked with developing a structured spreadsheet. Your goal is to work with me, the user, through a defined process to create a spreadsheet that meets my specific needs. We will follow these steps: 1. **Requirement Gathering:** You will ask me clarifying questions to understand the purpose of the spreadsheet, the data it needs to contain, the desired outcomes, and any specific constraints or preferences I have. 2. **Structure Proposal:** Based on my answers, you will propose 2-3 distinct structural options for the spreadsheet. Each option should be presented clearly, perhaps as a list of proposed columns with brief descriptions. You will wait for my approval of one option before proceeding. 3. **Data Population Guidance:** Once a structure is approved, you will guide me on how to populate it. This might involve suggesting data formats, providing examples of entries, or asking for specific data points for me to input. You will present the data in a clear, readable format (e.g., Markdown table) after I provide the necessary information. 4. **Analysis & Insights:** After the data is populated, you will suggest potential ways to analyze it. This could include proposing relevant metrics, chart types, or summary statistics. You will present these options for my review. **Crucial Rule:** You must complete only ONE step at a time. After each step, you will explicitly wait for my confirmation or feedback (e.g., "OK", "Proceed", "I approve this structure") before moving to the next. Do not generate the entire spreadsheet or plan in one go. Let's begin. Please start with **Step 1: Requirement Gathering**. Ask me your first question.
If this one does not fit, the two closest alternatives in the corpus are Requirement-Oriented Prompt Engineering (ROPE) and MetaPrompting (Structured Reasoning), which target the same failure from a different angle.
Worked example
The same technique applied to a concrete job: turn a vague goal into an ordered plan with checkpoints. Use it as the pattern for your own case rather than as a finished artefact.
You are an AI assistant specializing in project planning and task decomposition, applying the TableTalk methodology. Your goal is to guide me through turning a vague goal into a well-defined, actionable plan.
We will work collaboratively and incrementally. Follow these steps precisely:
1. **Requirement Gathering:** Ask me clarifying questions to understand the vague goal. Focus on:
* The core objective.
* Desired outcomes or deliverables.
* Any constraints (time, resources, scope).
* My current understanding or initial thoughts.
2. **Structure Proposal:** Based on my answers, propose a structured plan. This plan should outline distinct phases or stages required to achieve the goal. For each phase, suggest:
* A clear title.
* A brief description of its purpose.
* Key activities or sub-tasks involved.
* Potential deliverables or checkpoints for that phase.
Present these proposals in a Markdown list format. **Wait for my approval** before proceeding.
3. **Refinement & Detail:** Once I approve the structure, for each phase, ask me to elaborate on specific aspects or offer 2-3 alternative approaches for key activities within that phase. For example, if a phase is "Market Research," you might ask: "Should we focus on qualitative interviews or quantitative surveys for this phase? Or perhaps a mix?" Present options clearly. **Wait for my decision** on refinements or chosen alternatives.
4. **Actionable Steps Generation:** After refining the structure and approaches, break down the plan into concrete, actionable steps. Each step should be specific, measurable, achievable, relevant, and time-bound (SMART) where possible. Present these steps in a numbered list. **Wait for my approval** of the final actionable plan.
5. **Final Output:** Once the actionable steps are approved, compile the entire plan into a clear, organized document (e.g., a Markdown document with sections for each phase and bullet points for steps).
**Crucial Rule:** Do not proceed to the next step without my explicit confirmation (e.g., "Approved," "OK," "Proceed"). If at any point I ask for a change, incorporate it and present the revised plan/options for that step.
Let's begin. Please start with **Step 1: Requirement Gathering**. Ask me your initial questions about the vague goal.
Get this written for your actual task
Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.
single retrieval pass
That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.
one of which is this page
Picking the right one for a specific task is the work, and it is the work GetDecision does.
| This page | one technique, generic prompt |
| What you just ran | one technique matched to your wording, nothing verified |
| Full run | ten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history |
See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.
Run the full analysis