Tracking Status (Auto-Captured)

Auto-Track: ![ -f .claude/.ref-autotrack ] && echo "✓ Enabled" || echo "✗ Disabled" Sources Tracked: !wc -l < claude_usage/sources.md 2>/dev/null || echo "0" Prompts Tracked: !grep -c '^Prompt:' claude_usage/prompts.md 2>/dev/null || echo "0" Git Status: !git rev-parse --is-inside-work-tree &>/dev/null && echo "✓ Git repo" || echo "✗ Not a git repo"

Quick Example

/track:auto on
# ✓ Hooks-based tracking enabled
# PostToolUse: WebSearch/WebFetch/Read/Grep → claude_usage/sources.md
# SessionEnd: Major prompts → claude_usage/prompts.md

auto - Control Hooks-Based Auto-Tracking

Toggle or explicitly set automatic hooks-based tracking state.

Usage

/track:auto          # Toggle between enabled/disabled
/track:auto on       # Explicitly enable
/track:auto off      # Explicitly disable

What it does

1. Check current status:

   • Looks for ./.claude/.ref-autotrack marker file
   • If exists → hooks-based tracking is enabled
   • If missing → hooks are inactive

2. Update state:

   • No argument → toggle current state
   • Argument "on" → enable (create marker if missing)
   • Argument "off" → disable (delete marker if exists)

3. When enabling:

   • Creates ./.claude/.ref-autotrack with metadata:

     # Track Plugin v2.0 - Automatic Tracking Enabled
     #
     # Hooks configured:
     # - PostToolUse: Tracks WebSearch, WebFetch, Read, Grep
     # - UserPromptSubmit: Captures user prompts
     # - SessionEnd: Pairs prompts with outcomes
     #
     # Initialized: <timestamp>
     

   • Hooks activate automatically
   • No skill invocation needed

4. When disabling:

   • Deletes ./.claude/.ref-autotrack marker file
   • Hooks stop running
   • Temporary files cleaned up

5. Show new status:

   • Reports whether hooks-based tracking is now enabled or disabled
   • Shows what will be tracked based on current config
   • Explains how to change verbosity settings

Output examples

When enabling:

✓ Hooks-based tracking enabled

Automatic tracking via hooks:
- PostToolUse: Research sources (WebSearch/WebFetch/Read/Grep) → claude_usage/sources.md
- SessionEnd: Major prompts and outcomes → claude_usage/prompts.md

Current verbosity settings:
- Prompts: major (only significant academic/development work)
- Sources: all (every search operation)

Hooks run automatically - no manual intervention needed.

Use /track:config to adjust verbosity settings.

When disabling:

✓ Hooks-based tracking disabled

Hooks will no longer run automatically.

Tracked files remain intact:
- claude_usage/sources.md
- claude_usage/prompts.md

To re-enable: /track:auto on

When toggling:

✓ Hooks-based tracking toggled: OFF → ON

[Shows enabled message above]

Prerequisites

Run /track:init first to create .claude/ directory structure.

Notes

• State changes do not affect existing tracked data
• Configuration in .claude/.ref-config persists across state changes
• Marker file includes explanatory content and initialization timestamp
• Hooks activate/deactivate instantly based on marker presence
• Temporary files in .claude/.track-tmp/ are cleaned on disable

v2.0 Changes

Changed from v1.x:

• Old: Skill-based tracking (ref-tracker skill checked marker)
• New: Hooks-based tracking (hooks check marker)
• Benefit: Fully automatic, no skill activation needed
• Migration: Existing markers work unchanged

Implementation

Auto-tracking toggle functionality is implemented in scripts/auto.sh:

# Get skill directory
SKILL_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"

# Execute auto script with arguments
bash "$SKILL_DIR/scripts/auto.sh" "$@"

Script: skills/auto/scripts/auto.sh (102 lines)

Features:

• Argument parsing (on/off/toggle)
• State detection via marker file
• Marker creation with metadata
• Temporary file cleanup on disable
• Status reporting with verbosity settings
• Toggle state tracking

See scripts/auto.sh for full implementation details.