From vibe-explainer
Generate a visual "Change Report" after completing any coding task, explaining what changed in plain English with Mermaid diagrams and architecture maps. Use this skill every time you finish implementing, refactoring, fixing, or modifying code — before giving your final summary to the user. Also trigger when the user asks "what did you change?", "explain what happened", "show me a map of the changes", "what does my project look like now?", or any variation of wanting to understand their codebase or recent changes visually. This skill is designed for vibe coders who may not read code fluently and need to understand their project through diagrams and plain-language descriptions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/vibe-explainer:vibe-explainerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Generate a visual Change Report after every coding task so non-developers can understand what happened and how their project works.
Generate a visual Change Report after every coding task so non-developers can understand what happened and how their project works.
The person using this skill may not read code. They're building with AI and need to understand their project well enough to make decisions, explain it to others, debug problems, and stay in control. Your job is to be the translator between code and comprehension.
After completing any task that changes code, creates files, modifies database schemas, adds dependencies, or alters how components connect to each other. This includes bug fixes, new features, refactors, config changes, and dependency updates.
Also generate on demand when the user asks to understand their codebase or recent changes.
Before writing anything, assess what actually changed:
Not every change needs every section. Scale the report to the work:
Use the output template in references/output-template.md. Read it before writing.
Use the Mermaid pattern library in references/mermaid-patterns.md for diagram syntax. Read it before writing any diagrams.
These are non-negotiable:
No unexplained jargon. If you use a technical term, define it inline in parentheses the first time. Example: "a React component (a reusable building block of your user interface)".
Lead with what it does, not how it works. "Users can now reset their password from the login screen" comes before any mention of API routes or database columns.
Every diagram needs a plain-English caption explaining what the reader is looking at and why it matters.
File paths are always relative to the project root. No absolute paths.
Group related changes together by feature or purpose, not by file type. Don't list "all modified .tsx files" — instead group by "Login feature changes" or "Database updates".
Be honest about risk. If a change touches authentication, payments, data deletion, or anything else with real consequences, flag it clearly in a "Heads Up" callout.
Use the phrase "What this means" liberally. After any technical explanation, add a sentence starting with "What this means:" that translates it into practical impact.
Action items go at the end. If the user needs to do anything (set an env variable, run a migration, restart a server, test something manually), collect these into a clear checklist at the bottom.
Output the report as a markdown file saved to docs/change-reports/YYYY-MM-DD-HH-MM-description.md in the project directory. Use a short kebab-case description of the change, e.g., 2025-06-15-14-30-add-user-auth.md.
Also print a brief summary (3-5 sentences max) directly in the chat, ending with: "Full change report saved to docs/change-reports/[filename]."
If the docs/change-reports/ directory doesn't exist, create it.
When the user asks for a full project overview (not just a change report), generate a comprehensive map covering:
Save this as docs/PROJECT-MAP.md and tell the user where to find it.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub charliearlie/vibe-explainer --plugin vibe-explainer