skills.rest
ndestates
ndestatesCOMMUNITY
Explore all skills by this author →
Website𝕏@nickdodsleyGitHub

token-usage-meter

Monitors xAI Grok session token usage and prompt-cache hit rates with cost estimates.

Updated Jun 17, 2026
Search Tags:#xai#grok#context tracking#usage analysis#cost estimation#prompt caching#token monitoring
One-click install
npx skills add https://github.com/ndestates/ndestates-website --skill token-usage-meter-ndestates
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: token-usage-meter
Source: https://github.com/ndestates/ndestates-website/tree/main/.copilot/skills/token-usage-meter
Command: npx skills add https://github.com/ndestates/ndestates-website --skill token-usage-meter-ndestates

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Tracking token consumption and prompt-cache efficiency across Grok Build sessions is manual and error-prone, making it hard to detect context bloat, cache misses, and rising API costs before they degrade performance. ## Core Features & Use Cases - Session Context Monitoring: Syncs totalTokens from Grok updates.jsonl logs into reports/tokens/ and warns when context exceeds 100k tokens or a single turn grows by 15k tokens. - Prompt-Cache Analysis: Parses xAI API usage JSON to compute cache hit rates and flags POOR status when cached_tokens is zero on substantial prompts. - Cost Estimation: Applies per-model pricing from references/pricing.json to estimate session and per-turn spend, including cache savings versus no-cache baselines. - Use Case: During a long coding session, the agent appends a one-line Token meter footer to each response showing context size, turn delta, and estimated cost, and warns you to summarize the thread when context approaches 128k tokens. ## Quick Start Ask the agent to run the token monitor sync and show the current session's token usage, cache status, and estimated cost.

Frequently Asked Questions about token-usage-meter

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I monitor token usage in Grok Build sessions?▼

Run token_monitor.py with the --sync flag to parse the session's updates.jsonl file and log per-turn token counts to reports/tokens/. Use --footer to print a one-line summary showing context size, turn delta, and estimated cost.

How to check xAI API prompt cache hit rate?▼

Pipe the API response's usage JSON into analyze_token_usage.py, which reads cached_tokens from prompt_tokens_details and computes the hit rate as cached tokens divided by prompt tokens. Rates of 80% or higher are rated EXCELLENT, while zero cached tokens on prompts over 50 tokens is flagged POOR.

What token thresholds trigger context warnings?▼

Context at or above 100,000 tokens triggers a WARNING recommending cache-first lean replies, and 128,000 tokens triggers CRITICAL status suggesting thread summarization. A single-turn delta of 15,000 tokens or more also triggers a warning about large file reads.

Does the token monitor estimate API costs?▼

Yes, it estimates costs using per-model pricing from references/pricing.json, billing uncached input, cached input, and output tokens at their respective rates. Session mode assumes a 75% cache hit on prior context after the first turn, while API mode uses exact usage JSON for precise billing.

Why does the cache analysis report a POOR status?▼

A POOR status means cached_tokens was zero on a prompt larger than 50 tokens, indicating the prompt cache was not utilized. This typically happens when prompts change between calls or the cache window expired; see references/xai-prompt-caching.md for cache mechanics.

Related Skills

Explore More Software Engineering Skills→
ai-ecoverseai-ecoverse

xai-grok

Orchestrate Grok models for chat, search, image generation, and function calling via the xAI API.

Official
Advanced
MantisWareMantisWare

budget

Track token usage and pause agents when budget thresholds are reached.

Community
Advanced
ShunsukeHayashiShunsukeHayashi

daily-reporter

Collect Git activity and task lists into a markdown daily report.

Community
Intermediate
roanbrasilroanbrasil
5

llm-observability

Trace LLM calls and track token costs across LangSmith, LangFuse, and Phoenix.

Community
Advanced