SkillHub

tokenguard-pro

v1.0.0

Token cost optimizer for OpenClaw agents. Scan usage patterns, identify waste (excessive context, repeated queries, inefficient tool use), and get actionable recommendations to cut AI costs by 50-95%. Use when: monthly API spend exceeds $100, reviewing agent efficiency, or optimizing high-volume wor...

Sourced from ClawHub, Authored by sammy-the-bot

Installation

Please help me install the skill `tokenguard-pro` from SkillHub official store. npx skills add sammy-the-bot/tokenguard-pro

TokenGuard Pro 🛡️

Cut your AI token costs by 50-95% without sacrificing output quality.

TokenGuard Pro analyzes your OpenClaw usage patterns to identify waste, suggest optimizations, and project your savings. Built for users spending $100–$3,600/month on AI APIs.

What It Does

  • Scans your session logs for token consumption patterns
  • Identifies waste: oversized contexts, repeated queries, model mismatches, inefficient tool chains
  • Recommends specific optimizations with estimated savings
  • Generates a detailed cost-reduction report

Installation

npx clawhub install tokenguard-pro

Quick Start

# Run full analysis on recent usage
tokenguard-analyze

# Analyze last 7 days
tokenguard-analyze --days 7

# Focus on specific waste type
tokenguard-analyze --focus caching

# Export report as JSON
tokenguard-analyze --format json --output report.json

When to Use

✅ Use this skill when:

  • Monthly API costs exceed $100
  • You suspect context bloat in long conversations
  • Same queries repeat across sessions
  • You're using expensive models for simple tasks
  • Agents chain too many tool calls unnecessarily

❌ Don't use this skill when:

  • Costs are already minimal (<$50/month) — overhead not worth it
  • You need real-time budget enforcement — use API provider controls
  • Looking for fine-grained per-request pricing — check provider dashboards

Common Optimizations Found

Waste Type Typical Savings Fix Complexity
Oversized context 40-60% Easy
Wrong model choice 20-40% Easy
Repeated queries 15-30% Medium
Inefficient tool chains 10-25% Medium
Missing response caching 25-50% Medium

Example Report Output

╔══════════════════════════════════════════════════════════╗
║              TOKENGARD PRO ANALYSIS REPORT               ║
╠══════════════════════════════════════════════════════════╣
║  Period Analyzed: 14 days                                ║
║  Estimated Monthly Spend: $1,240                         ║
╠══════════════════════════════════════════════════════════╣
║  TOP ISSUES FOUND:                                       ║
║  1. GPT-4 used for 78% of simple classification tasks    ║
║     → Switch to GPT-3.5: SAVE ~$420/mo                   ║
║  2. Average context size 8.2K tokens (recommend: 2K)     ║
║     → Prune old messages: SAVE ~$310/mo                  ║
║  3. 23 repeated API calls without caching                ║
║     → Add response caching: SAVE ~$180/mo                ║
╠══════════════════════════════════════════════════════════╣
║  PROJECTED SAVINGS: $910/month (73% reduction)           ║
╚══════════════════════════════════════════════════════════╝

Limitations

  • Analyzes historical patterns, not real-time spend
  • Estimates based on standard pricing; actual rates may vary
  • Cannot modify your code — provides recommendations only
  • Requires access to OpenClaw logs for analysis

Pricing

$49 one-time purchase — pays for itself if it saves you just one month of excess spend.

Support

Questions? File an issue at github.com/appincubator/tokenguard-pro