
feedback
by yonatangross
The Complete AI Development Toolkit for Claude Code — 159 skills, 34 agents, 20 commands, 144 hooks. Production-ready patterns for FastAPI, React 19, LangGraph, security, and testing.
SKILL.md
name: feedback description: Manage the OrchestKit feedback system that learns from your usage. Use when providing feedback, reporting issues, suggesting improvements. context: inherit version: 1.2.0 author: OrchestKit tags: [feedback, learning, patterns, metrics, privacy, analytics, consent] user-invocable: true
Feedback - Manage Learning System
View and manage the OrchestKit feedback system that learns from your usage.
Overview
- Checking feedback system status
- Pausing/resuming learning
- Resetting learned patterns
- Exporting feedback data
- Managing privacy settings
- Enabling/disabling anonymous analytics sharing
- Viewing privacy policy
Usage
/feedback # Same as status
/feedback status # Show current state
/feedback pause # Pause learning
/feedback resume # Resume learning
/feedback reset # Clear learned patterns
/feedback export # Export feedback data
/feedback settings # Show/edit settings
/feedback opt-in # Enable anonymous sharing
/feedback opt-out # Disable anonymous sharing
/feedback privacy # View privacy policy
/feedback export-analytics # Export anonymous analytics for review
Subcommands
status (default)
Show the current feedback system state.
Output:
Feedback System Status
-----------------------------
Learning: Enabled
Anonymous sharing: Disabled
Data retention: 90 days
Learned Patterns:
- Auto-approves: npm install, npm test, git push (3 commands)
- Code style: async/await preferred, TypeScript strict mode
Agent Performance:
- backend-architect: 94% success (28 spawns) [improving]
- test-generator: 72% success (18 spawns) [declining]
Context Savings: ~8k tokens/session (estimated)
Storage: .claude/feedback/ (45 KB)
pause
Temporarily pause all learning without clearing data.
Action:
- Set
pausedUntilin preferences to a far future date - Confirm to user
Output:
Feedback learning paused
Your existing patterns are preserved.
Resume with: /feedback resume
resume
Resume paused learning.
Action:
- Clear
pausedUntilin preferences - Confirm to user
Output:
Feedback learning resumed
The system will continue learning from your usage.
reset
Clear all learned patterns (requires confirmation).
Action:
- Show what will be deleted
- Ask for confirmation (user must type "RESET")
- If confirmed, clear patterns file but keep preferences
Output (before confirmation):
WARNING: This will clear all learned patterns:
- 5 auto-approve permission rules
- 3 code style preferences
- Agent performance history
Your preferences (enabled, sharing, retention) will be kept.
To confirm, respond with exactly: RESET
To cancel, respond with anything else.
Output (after confirmation):
Feedback data reset
- Cleared 5 permission patterns
- Cleared 3 style preferences
- Cleared agent metrics
Learning will start fresh from now.
export
Export all feedback data to a JSON file.
Action:
- Read all feedback files
- Combine into single export
- Write to
.claude/feedback/export-{date}.json
Output:
Exported feedback data to:
.claude/feedback/export-2026-01-14.json
Contains:
- 5 learned permission patterns
- 3 code style preferences
- 8 skill usage metrics
- 4 agent performance records
File size: 12 KB
settings
Show current settings with option to change.
Output:
Feedback Settings
-----------------------------
enabled: true (master switch)
learnFromEdits: true (learn from code edits)
learnFromApprovals: true (learn from permissions)
learnFromAgentOutcomes: true (track agent success)
shareAnonymized: false (share anonymous stats)
retentionDays: 90 (data retention period)
To change a setting, use:
/feedback settings <key> <value>
Example:
/feedback settings retentionDays 30
opt-in
Enable anonymous analytics sharing. Records GDPR-compliant consent.
Action:
- Record consent in consent-log.json with timestamp and policy version
- Set shareAnonymized = true in preferences
- Confirm to user
Output:
Anonymous analytics sharing enabled.
What we share (anonymized):
- Skill usage counts and success rates
- Agent performance metrics
- Hook trigger counts
What we NEVER share:
- Your code or file contents
- Project names or paths
- Personal information
- mem0 memory data
Disable anytime: /feedback opt-out
opt-out
Disable anonymous analytics sharing. Revokes consent.
Action:
- Record revocation in consent-log.json with timestamp
- Set shareAnonymized = false in preferences
- Confirm to user
Output:
Anonymous analytics sharing disabled.
Your feedback data stays completely local.
No usage data is shared.
Re-enable anytime: /feedback opt-in
privacy
Display the full privacy policy for anonymous analytics.
Action:
- Display comprehensive privacy documentation
- Show what's collected, what's never collected
- Explain data protection measures
Output:
═══════════════════════════════════════════════════════════════════
ORCHESTKIT ANONYMOUS ANALYTICS PRIVACY POLICY
═══════════════════════════════════════════════════════════════════
WHAT WE COLLECT (only with your consent)
────────────────────────────────────────────────────────────────────
✓ Skill usage counts - e.g., "api-design used 45 times"
✓ Skill success rates - e.g., "92% success rate"
✓ Agent spawn counts - e.g., "backend-architect spawned 8 times"
✓ Agent success rates - e.g., "88% tasks completed successfully"
✓ Hook trigger counts - e.g., "git-branch-protection triggered 120 times"
✓ Hook block counts - e.g., "blocked 5 potentially unsafe commands"
✓ Plugin version - e.g., "4.12.0"
✓ Report date - e.g., "2026-01-14" (date only, no time)
WHAT WE NEVER COLLECT
────────────────────────────────────────────────────────────────────
✗ Your code or file contents
✗ Project names, paths, or directory structure
✗ User names, emails, or any personal information
✗ IP addresses (stripped at network layer)
✗ mem0 memory data or conversation history
✗ Architecture decisions or design documents
✗ API keys, tokens, or credentials
✗ Git history or commit messages
✗ Any data that could identify you or your projects
YOUR RIGHTS
────────────────────────────────────────────────────────────────────
• Opt-out anytime: /feedback opt-out
• View your data: /feedback export-analytics
• Check status: /feedback status
• View this policy: /feedback privacy
export-analytics
Export anonymous analytics data to a file for review before sharing.
Usage:
/feedback export-analytics [path]
If no path is provided, exports to .claude/feedback/analytics-exports/ with a timestamp.
Output:
Analytics exported to: .claude/feedback/analytics-exports/analytics-export-20260114-120000.json
Contents preview:
-----------------
Date: 2026-01-14
Plugin Version: 4.12.0
Summary:
Skills used: 8
Skill invocations: 45
Agents used: 3
Agent spawns: 12
Hooks configured: 5
Please review the exported file before sharing.
Export Format:
{
"timestamp": "2026-01-14",
"plugin_version": "4.12.0",
"skill_usage": {
"api-design-framework": { "uses": 12, "success_rate": 0.92 }
},
"agent_performance": {
"backend-system-architect": { "spawns": 8, "success_rate": 0.88 }
},
"hook_metrics": {
"git-branch-protection": { "triggered": 45, "blocked": 3 }
},
"summary": {
"unique_skills_used": 8,
"unique_agents_used": 3,
"hooks_configured": 5,
"total_skill_invocations": 45,
"total_agent_spawns": 12
},
"metadata": {
"exported_at": "2026-01-14T12:00:00Z",
"format_version": "1.0",
"note": "Review before sharing"
}
}
What We Share (When Opted In)
Anonymous analytics include only aggregated, non-identifiable data:
| Data Type | What's Included | What's NOT Included |
|---|---|---|
| Skills | Usage counts, success rates | File paths, code content |
| Agents | Spawn counts, success rates | Project names, decisions |
| Hooks | Trigger/block counts | Command content, paths |
What We NEVER Share
The following are blocked by design and never included in analytics:
- Project names, paths, or directory structure
- File contents, code, or diffs
- Decision content or context
- User identity, email, or credentials
- mem0 memory data
- URLs, IP addresses, or hostnames
- Any strings that could identify the project or user
Consent Management
Consent is managed per GDPR requirements:
- Explicit opt-in required - No data shared until you actively consent
- Audit trail - All consent actions logged in
consent-log.json - Easy revocation - Opt-out is as easy as opt-in
- Version tracking - Consent version tracked for policy changes
Security Note
The following commands are NEVER auto-approved regardless of learning:
rm -rf,sudo,chmod 777- Commands with
--forceor--no-verify - Commands involving passwords, secrets, or credentials
Related Skills
- skill-evolution: Evolve skills based on feedback
File Locations
See references/file-locations.md for storage details.
Analytics & Consent Files
.claude/feedback/
├── preferences.json # User preferences including shareAnonymized
├── consent-log.json # GDPR consent audit trail (opt-in/opt-out history)
├── metrics.json # Source data for analytics
└── analytics-exports/ # Exported analytics for review
└── analytics-export-*.json
Privacy Policy
See references/privacy-policy.md for full privacy documentation.
スコア
総合スコア
リポジトリの品質指標に基づく評価
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
1ヶ月以内に更新
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
レビュー
レビュー機能は近日公開予定です
