What it does
Reads through a batch of free-text feedback, reviews, survey answers, or support notes, and clusters it into a handful of recurring themes, each with a representative quote and an approximate count of how often it came up. Surfaces the single most common complaint and most common praise separately so nothing gets buried in a wall of text.
Install
This skill isn't packaged in an installable registry. It's a plain SKILL.md file you can copy directly. Copy it from the SKILL.md tab and save it at the path for your assistant:
Use it in your AI assistant
Claude Code
.claude/skills/customer-feedback-summarizer/ (or ~/.claude/skills/customer-feedback-summarizer/ for all projects)
GitHub Copilot (VS Code)
.github/skills/customer-feedback-summarizer/
OpenAI Codex
.codex/skills/customer-feedback-summarizer/
Cursor
.cursor/skills/customer-feedback-summarizer/ (or ~/.cursor/skills/customer-feedback-summarizer/ globally)
Compatibility
Any WordPress site; works from exported feedback text (reviews plugin export, survey CSV, support notes), no specific analytics integration required.
The complete SKILL.md for this skill:
---
name: customer-feedback-summarizer
description: Use when a user wants a batch of open-ended feedback or reviews summarized into themes, or asks "what are people saying" or for a "feedback roundup".
---
# Customer Feedback Summarizer
## When to use
A pile of free-text feedback, reviews, or survey answers has built up and the user wants the recurring themes surfaced rather than reading every entry themselves.
## Procedure
1. Read all feedback entries in the batch before grouping anything.
2. Cluster entries into 4-8 recurring themes based on subject matter, not sentiment alone.
3. For each theme, note an approximate count or percentage of entries and pick one representative quote.
4. Call out the single most common complaint and the single most common praise at the top of the summary, separate from the theme list.
5. Note any feedback that doesn't fit a theme cleanly rather than forcing it into the nearest bucket.
## Guardrails
- Counts are approximate groupings, not statistically validated, label them as such rather than presenting false precision.
- Don't discard a negative outlier just because it's rare, mention it if it describes a serious issue like a safety or billing problem.