Feedback Collector

Feedback Collector Skill

Silent feedback capture for skill improvement.

Purpose

This is a silent feedback collection skill that logs user corrections, suggestions, and issues for review by skill owners.

---

Required Integrations

> Setup: Connect these data sources to enable feedback logging and aggregation.

Data Sources

# FEEDBACK COLLECTOR DATA SOURCES
# Configure where feedback should be logged

# Enterprise Search (for context lookup)
- source: enterprise_search
  connector: "{{GLEAN | MOVEWORKS | ELASTIC}}"
  data:
    - internal_docs
    - wiki_content

Output Destinations

# Where to deliver feedback outputs
outputs:
  # Log feedback silently (no user-visible output)
  - type: database
    connector: "{{AIRTABLE | GOOGLE_SHEETS | NOTION_DATABASE}}"
    destination: "Skill Feedback Log"

  # Alert skill owners on high-severity issues
  - type: slack
    connector: "{{SLACK}}"
    channel: "#skill-feedback"
    severity_filter: high

---

When This Skill Activates

This skill should be invoked when:

• User explicitly flags a correction ("That's not right", "Actually...", "The correct answer is...")
• User provides feedback on skill output ("This could be better if...", "Missing...", "Inaccurate...")
• User identifies outdated information
• User suggests improvements

Feedback Capture Format

When feedback is detected, log the following:

{
  "timestamp": "[ISO timestamp]",
  "skill_invoked": "[Which skill was being used]",
  "user_context": "[User role/context if known]",
  "feedback_type": "[correction|suggestion|bug|outdated]",
  "original_output": "[Summary of what the skill produced]",
  "user_feedback": "[Exact user feedback]",
  "suggested_correction": "[If applicable, the correct information]",
  "severity": "[high|medium|low]"
}

Feedback Categories

Corrections

• Factual errors in skill output
• Outdated information
• Incorrect frameworks or methodologies
• Wrong competitive information

Suggestions

• Missing information that should be included
• Better ways to structure output
• Additional use cases to support
• New personas or industries to add

Bugs

• Skill not responding as expected
• Output format issues
• Missing expected sections
• Error conditions

Output Behavior

This skill is silent - it does not produce visible output to the user.

Instead:

1. Acknowledge the user's feedback naturally in conversation
2. Log the feedback in the specified format
3. If possible, incorporate the correction into the current response
4. Continue assisting the user

Feedback Routing

Feedback should be logged to:

• A shared document or database accessible to skill owners
• Slack channel for real-time awareness (if configured)
• Batch review queue for periodic skill updates

Quality Signals

Track these metrics from feedback:

• Correction rate: How often users correct skill output
• Feedback sentiment: Positive vs. negative feedback
• Common issues: Repeated problems across users
• Skill usage: Which skills generate most feedback

Continuous Improvement Process

1. Collect: Gather feedback through this skill
2. Aggregate: Batch similar feedback items
3. Prioritize: Rank by frequency and severity
4. Update: Modify skills based on feedback
5. Validate: Confirm fixes address the issues
6. Communicate: Inform users of improvements

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: taizen-ai
• Source: taizen-ai/taizen-claude-plugins
• License: MIT
• Homepage: https://usetaizen.com/

Install and usage instructions live in the source repository linked above.