/prune

Execute automatic context pruning to free up tokens immediately.

Auto-Pruning Analysis

Run a comprehensive pruning analysis using an agent to:

1. Scan Current Context

   • Identify all files currently loaded
   • Track when each file was last accessed
   • Determine which files are actively being used
   • Calculate token cost per file

2. Identify Pruning Candidates

   High Priority (Safe to Remove):

   • Files not accessed in last 30+ minutes
   • Test output files and logs
   • Temporary or cache files
   • Duplicate content
   • Archived/backup files accidentally loaded

   Medium Priority (Review Before Removal):

   • Files accessed 15-30 minutes ago
   • Large documentation files
   • Configuration files loaded for reference
   • Old tool outputs

   Low Priority (Keep):

   • Files accessed in last 15 minutes
   • Currently open/editing files
   • Core project files
   • Recent tool results

3. Calculate Savings For each candidate, estimate:

   • Token count
   • % of total context
   • Likelihood file is needed again
   • Risk score (data loss risk)

4. Execute Pruning

   Based on current usage level:

   If Critical (>85% usage):

   • Auto-remove high priority candidates
   • Ask permission for medium priority
   • Report total savings

   If Warning (60-85% usage):

   • Show pruning recommendations
   • Ask which to remove
   • Execute user choices

   If Healthy (<60% usage):

   • Show what could be pruned
   • Don't execute unless requested
   • Provide preventive guidance

5. Pruning Report

   ╔══════════════════════════════════════════════╗
   ║ CONTEXT PRUNING REPORT                       ║
   ╠══════════════════════════════════════════════╣
   ║ Files Analyzed: 25                           ║
   ║ Pruning Candidates: 8                        ║
   ║ Potential Savings: 45,000 tokens (22%)       ║
   ║                                              ║
   ║ AUTO-PRUNE ACTIONS:                          ║
   ║ ✓ Removed test-output.log (12K tokens)      ║
   ║ ✓ Removed old-schema.sql (18K tokens)       ║
   ║ ✓ Removed backup-config.json (8K tokens)    ║
   ║ ✓ Cleared old tool results (7K tokens)      ║
   ║                                              ║
   ║ Total Freed: 45,000 tokens                   ║
   ║ New Usage: 55% (down from 77%)               ║
   ║ Status: 🟢 Healthy                          ║
   ╚══════════════════════════════════════════════╝
   

Pruning Rules

Configure automatic pruning rules in your config:

{
  "pruning": {
    "enabled": true,
    "auto_prune_threshold": 70,
    "rules": {
      "remove_after_minutes": 30,
      "keep_patterns": ["src/**/*.js", "core/**"],
      "always_prune_patterns": ["*.log", "*.tmp", "*backup*"],
      "min_file_size_kb": 5
    }
  }
}

Smart Pruning Features

• Access Tracking: Monitors file access timestamps
• Usage Prediction: ML-based prediction of file reuse likelihood
• Dependency Analysis: Keeps files referenced by active code
• Undo Buffer: Can restore recently pruned files
• Pruning History: Track what was pruned and savings achieved

Execution Options

Option 1: Automatic (Recommended for Critical)

Please auto-prune my context immediately. I'm at critical usage.

Option 2: Interactive (Recommended for Warning)

Show me what can be pruned and let me choose.

Option 3: Dry Run (Recommended for Healthy)

Show me what would be pruned but don't execute.

Option 4: Aggressive (Emergency Only)

Aggressively prune everything not used in last 15 minutes.

Please specify your preference or let me choose based on current usage level.