Skip to content

hoavdc/CodexKit / skills/codexkit-csat-sentiment-analyzer

codexkit-csat-sentiment-analyzer

Unverified

Analyze CSAT, NPS comments, reviews, and support feedback for sentiment, recurring themes, customer pain, and service improvement actions. Use for customer support and CX feedback reviews.

CSAT Sentiment Analyzer

When to Use

  • Analyzing CSAT, NPS, app reviews, support comments, or post-interaction feedback.
  • Finding recurring customer pain points and service improvement themes.
  • Preparing support, CX, product, or leadership feedback summaries.
  • Comparing sentiment across segments, channels, agents, products, or time periods.

Procedure

Step 1 - Normalize Feedback

Identify source, date range, channel, score type, segment, and any metadata. Keep raw counts separate from percentages.

Step 2 - Classify Sentiment

Use a simple sentiment label:

  • positive
  • neutral
  • negative
  • mixed
  • unclear

Include confidence when comments are short or ambiguous.

Step 3 - Code Themes

Group feedback into themes such as speed, quality, pricing, reliability, usability, billing, support tone, missing features, or documentation.

Step 4 - Quantify Patterns

Report counts, percentages, average score, trend direction, and representative examples. Avoid claiming statistical significance without enough data.

Step 5 - Recommend Actions

Link every action to a theme and owner group: support, product, docs, billing, success, operations, or leadership.

Inputs

InputRequiredFormat
Feedback datasetYesComments, scores, reviews, tickets
Date rangeRecommendedStart and end date
SegmentsOptionalPlan, region, product, channel, agent
Scoring systemOptionalCSAT 1-5, NPS, thumbs up/down
Business contextOptionalLaunch, outage, policy change

Output

## CSAT Sentiment Analysis - [Period]

### Executive Summary
[Key trend, sentiment, and action]

### Score Snapshot
| Metric | Value | Notes |
|--------|-------|-------|

### Theme Breakdown
| Theme | Sentiment | Count | Percent | Representative Comment | Recommended Action |
|-------|-----------|-------|---------|------------------------|--------------------|

### Segment Differences
| Segment | Pattern | Confidence |
|---------|---------|------------|

### Action Plan
| Owner | Action | Evidence | Priority |
|-------|--------|----------|----------|

Quality Criteria

  • Counts and percentages reconcile to the input size.
  • Themes are mutually understandable and not over-fragmented.
  • Representative comments are short and anonymized when needed.
  • Recommendations are tied to evidence.
  • Sensitive customer data is removed or masked.

Verification (4C)

CheckQuestion
CorrectnessDo sentiment labels and theme counts match the raw feedback?
CompletenessAre scores, themes, segments, evidence, and actions included?
Context-fitAre the recommendations realistic for the support or CX team?
ConsequenceCould a small sample or biased feedback source lead to the wrong product or staffing decision?

Edge Cases

  • Very small sample - Mark findings as directional and avoid percentages that imply precision.
  • Sarcasm or mixed sentiment - Use "mixed" or "unclear" instead of forcing polarity.
  • Personally identifiable information - Mask names, emails, phone numbers, and account IDs.
  • Outage or one-off incident - Separate incident-driven feedback from baseline sentiment.

Examples

Prompt: "Analyze these 300 CSAT comments from April. Break down sentiment, top pain themes, and the three actions support should take next."

Good pattern: "Billing confusion appears in 42 of 300 comments (14%). The most common evidence is customers not understanding prorated invoices after plan changes."

Definition of Done

  • Sentiment and theme counts are transparent.
  • Representative examples support the themes.
  • Actions are owner-specific and prioritized.
  • Bias and sample limits are documented.

Changelog

  • v1.0.0 - Initial release
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 464kUpdated Claude CodeCodexVerifiedInstall
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 464kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 318kUpdated Claude CodeCodexUnverifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 318kUpdated Claude CodeCodexUnverifiedInstall