/ork:feedback                    # Same as status
/ork:feedback status             # Show current state
/ork:feedback pause              # Pause learning
/ork:feedback resume             # Resume learning
/ork:feedback reset              # Clear learned patterns
/ork:feedback export             # Export feedback data
/ork:feedback settings           # Show/edit settings
/ork:feedback opt-in             # Enable anonymous sharing
/ork:feedback opt-out            # Disable anonymous sharing
/ork:feedback privacy            # View privacy policy
/ork:feedback export-analytics   # Export anonymous analytics for review
/ork:feedback bug                # File a bug report as a GitHub issue
/ork:feedback bug [description]  # File with pre-filled description

# Pre-push hook auto-approves all commands
git push --force origin main  # DANGEROUS
rm -rf /important/directory    # DESTRUCTIVE

# Security check in hook
if [[ "$cmd" =~ (rm -rf|sudo|chmod 777|--force|--no-verify) ]]; then
  echo "BLOCKED: Command requires manual approval"
  exit 1
fi

AskUserQuestion(questions=[
  {
    "question": "What went wrong?",
    "header": "Bug Description",
    "options": [
      {"label": "Skill issue", "description": "A skill (/ork:*) didn't work as expected"},
      {"label": "Agent issue", "description": "A spawned agent produced wrong results or failed"},
      {"label": "Hook issue", "description": "A hook blocked something it shouldn't, or missed something"},
      {"label": "Build/install issue", "description": "Plugin build, installation, or setup problem"},
      {"label": "Other", "description": "Something else went wrong"}
    ],
    "multiSelect": false
  }
])

# Version from package.json
Read(file_path="{project_dir}/package.json")  # extract "version" field

# OS info
Bash(command="uname -s -r")

# Claude Code version (if available)
Bash(command="claude --version 2>/dev/null || echo 'unknown'")

# Git branch
Bash(command="git branch --show-current 2>/dev/null || echo 'unknown'")

gh issue list -R yonatangross/orchestkit \
  --search "{keywords from description}" \
  --state open \
  --json number,title,url \
  --limit 5

gh issue create -R yonatangross/orchestkit \
  --title "Bug: {concise_title}" \
  --label "bug" \
  --body "$(cat <<'EOF'
## Bug Report

**Description:** {user_description}

**Category:** {skill|agent|hook|build|other}

## Steps to Reproduce

{user_provided_steps_or "1. [Please describe steps to reproduce]"}

## Expected Behavior

{user_provided_or "N/A"}

## Actual Behavior

{user_description_of_what_happened}

## Environment

| Field | Value |
|-------|-------|
| OrchestKit | {version} |
| Claude Code | {cc_version} |
| OS | {os_info} |
| Branch | {branch} |

## Context

{if metrics available:}
- Last skill: `{skill_name}` ({success_rate}% success rate)
- Last agent: `{agent_name}` ({success_rate}% success rate)

---
*Filed via `/ork:feedback bug`*
EOF
)"

Bug report filed!

  Issue: #123
  URL:   https://github.com/yonatangross/orchestkit/issues/123
  Title: Bug: {title}

Track progress or add details at the URL above.

Could not create GitHub issue (gh CLI not authenticated).

Bug report saved to:
  .claude/feedback/pending-bugs/bug-20260226-143000.md

To file manually:
  1. Run: gh auth login
  2. Then: gh issue create -R yonatangross/orchestkit --body-file .claude/feedback/pending-bugs/bug-20260226-143000.md --title "Bug: {title}" --label bug

.claude/feedback/
├── preferences.json       # User preferences and settings
├── metrics.json           # Skill and agent usage metrics
├── learned-patterns.json  # Auto-approve patterns and code style
├── satisfaction.json      # Session satisfaction tracking
├── consent-log.json       # GDPR consent audit trail (#59)
├── analytics-sent.json    # Transmission history (if enabled)
├── analytics-exports/     # Anonymous analytics exports for review
│   └── analytics-export-*.json
└── pending-bugs/          # Bug reports saved when gh CLI unavailable
    └── bug-*.md

{
  "version": "1.0",
  "enabled": true,
  "learnFromEdits": true,
  "learnFromApprovals": true,
  "learnFromAgentOutcomes": true,
  "shareAnonymized": false,
  "consentedAt": null,
  "consentVersion": null,
  "syncGlobalPatterns": true,
  "retentionDays": 90,
  "pausedUntil": null
}

{
  "version": "1.0",
  "events": [
    {
      "action": "granted",
      "version": "1.0",
      "timestamp": "2026-01-14T10:00:00Z"
    },
    {
      "action": "revoked",
      "timestamp": "2026-01-15T14:00:00Z"
    }
  ]
}

{
  "version": "1.0",
  "updated": "2026-01-14T10:00:00Z",
  "skills": { ... },
  "hooks": { ... },
  "agents": { ... }
}

{
  "version": "1.0",
  "updated": "2026-01-14T10:00:00Z",
  "permissions": { ... },
  "codeStyle": { ... }
}

{
  "timestamp": "2026-01-14",
  "plugin_version": "4.12.0",
  "skill_usage": {
    "<skill-id>": { "uses": 12, "success_rate": 0.92 }
  },
  "agent_performance": {
    "<agent-id>": { "spawns": 8, "success_rate": 0.88 }
  },
  "hook_metrics": {
    "<hook-name>": { "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"
  }
}

.claude/feedback/

~/.claude/global-patterns.json

.claude/scripts/consent-manager.sh

.claude/scripts/analytics-lib.sh

.claude/scripts/analytics-sender.sh

.claude/schemas/
├── feedback.schema.json   # Metrics, patterns, preferences schemas
└── consent.schema.json    # Consent log schema (#59)

# Blocked patterns (excerpt from analytics-lib.sh)
/Users/, /home/, /var/, /tmp/    # File paths
@                                 # Email indicators
http://, https://                 # URLs
password, secret, token, api_key  # Credentials
username, user_id, email          # Identifiers
[0-9]{1,3}\.[0-9]{1,3}...        # IP addresses

# Via command
/ork:feedback opt-in

# What happens:
# 1. consent-log.json records: {"action": "granted", "version": "1.0", "timestamp": "..."}
# 2. preferences.json sets: shareAnonymized = true

# Via command
/ork:feedback opt-out

# What happens:
# 1. consent-log.json records: {"action": "revoked", "timestamp": "..."}
# 2. preferences.json sets: shareAnonymized = false
# 3. No further data collection occurs

/ork:feedback status
# Shows: consent state, timestamp, policy version

Session Activity
      │
      ▼
┌──────────────┐
│ Local Metrics│ ← Stored in .claude/feedback/metrics.json
│  (always)    │
└──────┬───────┘
       │
       ▼
┌──────────────────┐
│ Consent Check    │ ← has_consent() must return true
└──────┬───────────┘
       │
       ▼ (only if consented)
┌──────────────────┐
│ PII Validation   │ ← validate_no_pii() scans all data
└──────┬───────────┘
       │
       ├── PII Found → ABORT
       │
       └── Clean → Export/Send

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)

Feedback learning paused

Your existing patterns are preserved.
Resume with: /ork:feedback resume

Feedback learning resumed

The system will continue learning from your usage.

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.

Feedback data reset

- Cleared 5 permission patterns
- Cleared 3 style preferences
- Cleared agent metrics

Learning will start fresh from now.

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

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:
  /ork:feedback settings <key> <value>

Example:
  /ork:feedback settings retentionDays 30

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
  - Memory data

Disable anytime: /ork:feedback opt-out

Anonymous analytics sharing disabled.

Your feedback data stays completely local.
No usage data is shared.

Re-enable anytime: /ork:feedback opt-in

═══════════════════════════════════════════════════════════════════
              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)
  ✗ 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:     /ork:feedback opt-out
  • View your data:      /ork:feedback export-analytics
  • Check status:        /ork:feedback status
  • View this policy:    /ork:feedback privacy

/ork:feedback bug
/ork:feedback bug Something broke when I ran /ork:verify

Bug report filed!

  Issue: #456
  URL:   https://github.com/yonatangross/orchestkit/issues/456
  Title: Bug: verify skill fails on monorepo with pnpm workspaces

Track progress or add details at the URL above.

Could not create GitHub issue (gh CLI not authenticated).

Bug report saved to:
  .claude/feedback/pending-bugs/bug-20260226-143000.md

To file manually:
  1. Run: gh auth login
  2. Then: gh issue create -R yonatangross/orchestkit \
       --body-file .claude/feedback/pending-bugs/bug-20260226-143000.md \
       --title "Bug: ..." --label bug

/ork:feedback export-analytics [path]

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.

{
  "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"
  }
}