AI Skill Report Card

Elaborating Prompts Before Execution

Expands terse user input into a detailed, structured prompt (with goal, context, constraints, and workflow steps) and presents it for approval before doing any work. Use when a user gives a short or ambiguous request and would benefit from confirming a fully-specified plan before execution, or explicitly asks for prompt elaboration prior to acting.

B+78·Sep 18, 2026·Source: Web
13 / 15

On receiving any user input, do NOT execute the underlying task yet. Instead:

  1. Expand the input into an elaborated prompt (see template below).
  2. Show it to the user.
  3. Ask: "Proceed with this? (yes/no)"
  4. Only execute the task if the user replies affirmatively (yes, y, go ahead, confirmed, etc.).
  5. If the user says no, ask what to change and re-elaborate.

Example trigger: User says "write a landing page for my app" → Skill produces elaborated prompt → waits for "yes" → only then generates the landing page.

Recommendation
Add an example of a moderately complex or ambiguous request (not just a trivial one) to show how elaboration scales with complexity
13 / 15

Progress:

  • Step 1: Parse raw user input for intent, domain, and implicit requirements
  • Step 2: Fill the elaborated prompt template
  • Step 3: Present elaborated prompt to user, clearly separated from any commentary
  • Step 4: Wait for explicit approval
  • Step 5: If approved, execute the task per the elaborated prompt
  • Step 6: If rejected/modified, incorporate feedback and repeat from Step 2

Elaborated Prompt Template

Recommendation
Address edge cases like multi-turn back-and-forth elaboration, or when the user gives feedback that only partially changes the request

Goal: [one-sentence restatement of what the user wants]

Context: [relevant background inferred or asked for — platform, audience, constraints]

Requirements:

  • [explicit requirement 1]
  • [explicit requirement 2]
  • [implicit requirement inferred from phrasing]

Constraints/Assumptions:

  • [tech stack, tone, length, format, platform: Claude/Copilot/ChatGPT/etc.]
  • [any assumption made to fill gaps — flag clearly as an assumption]

Proposed Workflow:

  1. [step]
  2. [step]
  3. [step]

Deliverable: [exact form of output — file, code block, doc, list, etc.]


Proceed with this? (yes/no)

14 / 20

Example 1: Input: "write a landing page for my app" Output:

Recommendation
Clarify how this skill interacts with genuinely urgent/simple one-off requests where elaboration overhead may be undesirable — give guidance on when NOT to apply this skill or how to keep it lightweight

Goal: Create a marketing landing page for the user's app.

Context: No product name, audience, or platform specified — assuming a generic SaaS app targeting general users, deployed as a static HTML page.

Requirements:

  • Hero section with headline and CTA
  • Feature highlights (assumed 3)
  • Footer with basic links

Constraints/Assumptions:

  • Assuming plain HTML/CSS output (no framework specified)
  • Assuming English, professional tone

Proposed Workflow:

  1. Draft headline, subheadline, and CTA copy
  2. Outline feature section content
  3. Generate HTML/CSS structure
  4. Review for responsiveness and clarity

Deliverable: A single HTML file with embedded CSS


Proceed with this? (yes/no)


**Example 2:**
Input: "yes"
Output: Executes the previously approved elaborated prompt's workflow and returns the deliverable — no further elaboration.

**Example 3:**
Input: "no, make it for a fitness app instead"
Output: Re-elaborates the prompt substituting "fitness app" context, then re-asks for approval.
  • Always separate the elaborated prompt visually (heading, horizontal rule) from other text.
  • Flag assumptions explicitly instead of silently guessing.
  • Keep the elaborated prompt tight — no more than ~10 bullet lines total unless task complexity demands more.
  • Recognize varied affirmative replies ("yes", "sure", "sounds good", "ship it") as approval.
  • If the user's next message is unrelated to a pending elaborated prompt, treat it as a new input and elaborate fresh.
  • Carry forward context from rejected iterations rather than starting blank.
  • Do NOT execute the task in the same turn the elaborated prompt is shown.
  • Do NOT skip elaboration for "simple-looking" requests — always elaborate first.
  • Do NOT treat silence or unrelated replies as approval.
  • Do NOT bury the yes/no question inside paragraphs — always end with a clear, standalone confirmation prompt.
  • Do NOT re-elaborate from scratch after approval; execute directly using the already-approved prompt.
0
Grade B+AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
13/15
Workflow
13/15
Examples
14/20
Completeness
11/20
Format
14/15
Conciseness
13/15