agent-expenditure

This Skill addresses the challenge of quantifying token waste in multi-agent parallel scheduling. In practical development, when multiple AI agents are dispatched simultaneously to perform tasks, coordination overhead (such as file read/write conflicts) increases with the number of agents, leading to disproportionate token consumption relative to output. This tool tracks token usage for each agent, automatically identifies waste patterns, and helps teams avoid blindly increasing the agent count.

Usage is straightforward: you simply run this Skill after completing a parallel scheduling operation. It automatically analyzes each agent's token consumption, file conflict ratios, and generates a Brooks' Law comparison table. You can then query for waste signals using natural language instructions, such as "Check if there was duplicate work in this scheduling" or "Evaluate whether the number of agents should be reduced." It also provides a post-scheduling checklist to guide you in assessing the output value of each agent individually.

It is particularly suitable for teams or individuals who frequently engage in multi-agent parallel development, especially those already experiencing token budget overruns, duplicate agent outputs, or excessive coordination costs. The tool is not applicable to single-agent workflows or agents currently executing, as it focuses on post-hoc analysis.

It is recommended to use this tool for review after each large-scale parallel scheduling operation. Note that it primarily analyzes coordination overhead (e.g., file conflicts), not session-level token budget management. For budget control, it should be used in conjunction with the token-conservation skill. Additionally, this skill relies on the Claude Code plugin environment and requires prior installation.