Cost Tracker

What it does

Cost Tracker gives agents visibility into session token and dollar costs, surfaces top cost drivers, and recommends concrete optimizations (model choice, batching, selective reads). It provides budget checkpoints and a formatted output that helps operators decide when to compact, wrap up, or switch strategies.

When to use it

Use during interactive coding or multi-step agent workflows when you need to: check current session spend mid-session, set soft budget alerts, compare cost by task type (bug fix vs feature), or plan work across multiple sessions to avoid surprises. Valuable for long-running explorations and teams sharing compute budgets.

What's included

• Scripts: none bundled (has_scripts=false)
• References: none bundled (has_references=false)
• Instructions: The skill details how to read session cost metrics, interpret token budget tables, identify high-cost operations (large file reads, subagent spawn, broad searches), and apply optimizations like prompt scoping, model selection, and batching.

Compatible agents

Compatible with Claude Code and other agent runtimes that expose session token/cost metrics and support subagent workflows; generally useful across Claude, Codex, and Cursor-style agents.