improve-prompt

Validate translated prompts against originals and flag inconsistencies.

282|23|Updated Jan 3, 2026
One-click install
npx skills add https://github.com/team-attention/stanford-cs146s-kr --skill improve-prompt
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: improve-prompt
Source: https://github.com/team-attention/stanford-cs146s-kr/tree/main/.claude/skills/improve-prompt
Command: npx skills add https://github.com/team-attention/stanford-cs146s-kr --skill improve-prompt

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill validates translated prompts against the original sources, detects inconsistencies, and coordinates a fact-checking workflow to ensure high-quality prompt content across chapters and pages.

Core Features & Use Cases

  • Automated evaluation: Compare translated prompts with originals to identify concept drift and terminology mismatches.
  • Fact-check integration: Trigger WebSearch-based verification of extracted claims to ensure accuracy of technical terms and data.
  • Auditable remediation: Generate an issue-driven report, allow user confirmation, and apply approved fixes to prompts.
  • Use Case: If you publish translated course prompts, use this Skill to maintain alignment with the English source while documenting changes for review.

Quick Start

Run /improve-prompt week1/deep-dive-llms/tokenization to start a single-chapter evaluation. For a broader sweep, run /improve-prompt week1/deep-dive-llms --all-chapters. To generate a report only without applying changes, use /improve-prompt week1/slug --skip-apply.

Frequently Asked Questions about improve-prompt

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate translated prompts against the original source?

To validate translated prompts, this Skill compares them against originals to identify concept drift and terminology mismatches, then generates an issue-driven report for tracking and remediation.

How does fact-checking work for prompt evaluation?

Fact-checking for prompt evaluation triggers a WebSearch-based workflow to verify extracted claims, ensuring the accuracy of technical terms and data within your translated content.

Can I apply translation fixes automatically without manual review?

You cannot apply fixes automatically without review; this Skill generates an issue-driven report for user confirmation, then applies only the approved fixes to your prompts.

What is the best way to run prompt QA across multiple chapters?

The best way to run prompt QA across multiple chapters is using the broad sweep command with the --all-chapters flag to evaluate an entire directory structure at once.

Can I generate a prompt validation report without modifying the original files?

Yes, you can generate a validation report without modifying files by running the evaluation command with the --skip-apply flag to audit issues without applying changes.