AI Prompt Version Snapshot Card

Purpose

Use this skill when a user keeps revising a prompt for writing, analysis, coding help, image generation, tutoring, planning, or another AI workflow and wants to preserve the version that worked. The output is a copyable and printable prompt snapshot card with version labels, changed instructions, result notes, quality scores, privacy checks, and a keep or revise decision.

This is a prompt-only documentation artifact. It does not call AI tools, run prompts, store credentials, save private data, or guarantee output quality.

Use This Skill When

Use this skill when the user asks to:

• Track several versions of an AI prompt.
• Compare prompt changes and output results.
• Save the best prompt before it gets lost in chat history.
• Create a reusable prompt log, prompt testing template, or prompt iteration card.
• Record a privacy check before reusing or sharing prompt text.

Best Inputs

Ask only for the details needed to build the card. If the user does not know an answer, proceed with clear placeholders.

• Prompt task or workflow, such as writing, analysis, coding help, image generation, tutoring, or planning.
• What a successful output should look like.
• Non-sensitive prompt versions or summaries of the changed lines.
• Optional model or tool name, only if the user provides it.
• Test input summary and result notes.
• Any known limits, privacy concerns, or reuse rules.

Do not ask for secrets, credentials, confidential records, regulated data, client data, or unnecessary personal information. Tell the user to redact sensitive content before including prompt text.

Workflow

1. Identify the prompt task, target output, audience, and reuse context.
2. Ask the user to paste or summarize only non-sensitive prompt versions.
3. Create a version table with version label, changed line, model or tool name if provided, test input, result notes, and decision.
4. Add a quality score row for clarity, accuracy, format fit, tone fit, completeness, and effort saved.
5. Add a privacy and data-minimization check before any prompt is saved or reused.
6. Generate a best-version snapshot card that preserves the winning instruction set and known limitations.
7. Add a retire lane for confusing, risky, overlong, stale, or low-performing prompt versions.
8. Produce a next-test plan with one small change to try instead of rewriting everything.

Output Format

Return the artifact set in this order:

1. Prompt Task Snapshot

• Task:
• Desired output:
• Intended user or audience:
• Reuse context:
• Sensitive content removed:
• Assumptions:

2. Prompt Version Table

Create a table or list with these fields for each version:

• Version label
• Changed instruction or changed line
• Model or tool name, if user-provided
• Test input summary
• Result notes
• Keep, revise, or retire decision

3. Quality Score Row

Score each version from 1 to 5 for:

• Clarity
• Accuracy
• Format fit
• Tone fit
• Completeness
• Effort saved

4. Privacy Check

Include short checkboxes for:

• Private names removed or replaced
• Client or workplace details removed
• Regulated or sensitive data removed
• Credentials and account details absent
• Only needed context included
• Sharing or reuse rules checked

5. Best-Version Snapshot Card

Provide a compact card with:

• Best version name
• Goal
• Prompt text or safe summary
• Must-keep instruction
• Known limitation
• Best test input
• Best observed result
• Reuse note

6. Retire Lane

List versions to stop using, with one reason for each.

7. Next-Test Plan

Recommend one small change to test next and one thing to hold constant.

Style Rules

• Keep labels short enough for a printable card, sticky note, document sidebar, or shared workspace block.
• Use plain action words such as Keep, Revise, Retire, Test Next, and Redact First.
• Prefer placeholders such as [VERSION], [GOAL], [TEST INPUT], and [LIMITATION].
• Make the winning version easy to copy without mixing it with discarded versions.
• Keep uncertainty visible. Do not turn weak result notes into confident claims.

Safety Boundary

• Do not run prompts, call AI tools, automate tests, connect to accounts, or access chat histories.
• Do not request secrets, credentials, private keys, account data, confidential client data, or regulated personal data.
• Do not guarantee that a prompt will produce accurate, safe, legal, compliant, or high-quality output.
• Do not preserve sensitive prompt text when a summary or placeholder would work.
• Remind the user to remove private, confidential, regulated, or client data before saving or reusing prompt text.

Example Prompts

• "Help me compare three versions of this writing prompt and save the best one."
• "Make a prompt iteration card for my coding assistant prompt."
• "I finally got a good AI result. Create a snapshot card so I can reuse the prompt safely."
• "Turn these prompt changes into a version table with a privacy check."