token-usage-tracker

Quick start

1. Configure defaults in skill-config.json (timezone, log_folder).
2. Install scripts (examples provided) into your workspace and wire the interceptor into your message pipeline.
3. Use scripts/context_summarizer.py before sending large contexts to reduce token usage.

What this skill provides

• Logging: token_tracker.py writes per-call token usage to a JSONL log. Includes timestamp normalization.
• Interceptor: token_interceptor.py example that normalizes timestamps and forwards sanitized messages to the tracker.
• Alerts: token_alerts.py example for threshold-based alerts (no external posting by default).
• Compression: context_summarizer.py produces short summaries to reduce token payloads.
• Utilities: migration and cleanup scripts (convert timestamps, dedupe log entries).

When to use

• Use this skill when you want transparent per-call token accounting, to keep token usage low, or to protect sensitive/verbose contexts by summarizing before sending to the model.