arm-pz

Diagnose Usage — Development skill for Claude Code

Development community

Claude Code skill that works out why your usage limit drained fast.

How to install Diagnose Usage

This entry records only its repository, not the path inside it, so there is no exact command to give. Open arm-pz/diagnose-usage and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Diagnose Usage does

Claude Code skill that works out why your usage limit drained fast. Reads your /usage paste, checks the heavy-session causes Anthropic named (effort level, subagents, /goal, workflows, agent teams, SDK jobs) and ranks up to three culprits with one fix each.

Alternatives in Development

Analyze Feature Requests — Prioritize feature requests by theme, impact, effort, and risk 7.8k ★
Claude Token Efficient — One CLAUDE.md file 3.8k ★
Agentlas Call — Prepare explicitly named Agentlas Hub or Cloud agents 1.1k ★

README

diagnose-usage

An AI agent skill (a `SKILL.md` for Claude Code) that works out why your Claude Code usage limit drained faster than expected. You paste your `/usage` output; it checks it against the heavy-session causes Boris Cherny of the Claude Code team named on 29 Oct 2026 (a raised effort level, subagents inheriting it, `/goal`, workflows, Agent Teams, long SDK jobs) and gives you up to three ranked culprits with one change each. It does not claim a bug: Anthropic's review found no single root cause.

Siblings: [choose-effort](https://github.com/arm-pz/choose-effort) for deciding effort *before* a task. For plan limits, prices and API billing, see Anthropic's docs; this skill does not cover them.

Install

npx skills add arm-pz/diagnose-usage --skill diagnose-usage

Or copy the repo files into `~/.claude/skills/diagnose-usage/` and start a new session.

Ask it things like

"Why is my Claude Code usage draining so fast today?"
"I hit my Claude Code limit in an hour."
"What does /usage actually show me?"

Status

Structure checks done by hand; trigger cases in `references/trigger-cases.md` are hand-read against the skill descriptions and **not** run in a live session. `check_skill.py`: OK, 0 fail 0 warn. Effort levels and commands change often, so the skill tells the agent to confirm them against your `/effort` slider and the current docs.