Claude Usage Analyzer

Analyze Claude Code token usage. Shows where tokens went, which projects cost most, and how to reduce waste. Use when user asks about token usage, costs, or...

Security Scan

The skill's actions (parsing local Claude session logs with the ccwhy CLI) match its description and requested access, with no unexplained credentials or installs — but it relies on a third-party CLI you should vet.

Purpose & Capability

Name/description (analyze Claude Code token usage) aligns with the declared config path (~/.claude/projects/*/*.jsonl) and the instructions to run the ccwhy CLI to produce reports. No unrelated credentials or system access are requested.

Instruction Scope

Instructions are narrowly scoped to reading local Claude session logs and invoking ccwhy. The SKILL.md states 'offline, no network', but the agent's behavior depends on the external ccwhy binary (which could contact network services); the skill itself does not supply or install ccwhy.

Install Mechanism

Instruction-only skill with no install spec or shipped code — lowest install risk. It requires the user to pre-install the third-party ccwhy CLI (link provided), which is out-of-band and should be reviewed by the user.

Credentials

No environment variables, credentials, or unrelated config paths are requested. Access to ~/.claude/... logs is proportional to the stated purpose, but those logs may contain sensitive content and should be considered private.

Persistence & Privilege

No persistent or elevated privileges requested (always:false, no config writes, no system-wide changes). The skill is user-invocable only and does not autonomously require extra privileges.

Assessment

This skill simply parses your local Claude session logs using the ccwhy CLI. Before installing/using it: (1) confirm you have the exact logs at ~/.claude/projects/*/*.jsonl and are comfortable with them being read; (2) review and install ccwhy from the linked GitHub repo and audit its code or run it in a network-isolated environment if you need assurance it won't phone home; (3) consider redacting sensitive content from logs or backing them up first. If you don't trust the external ccwhy binary, do not run it — the SKILL.md does not bundle or vet that tool.

Analyze your Claude Code token usage by parsing local session logs.

Data access

• Reads ~/.claude/projects/*/*.jsonl (local Claude Code session logs)
• Runs offline, no network access, no API keys, no credentials
• Open source: https://github.com/SingggggYee/ccwhy

Usage

Requires the ccwhy CLI to be pre-installed. See https://github.com/SingggggYee/ccwhy for installation instructions.

ccwhy

More commands:

• ccwhy report --days 7 (last 7 days)
• ccwhy sessions (top sessions by cost)
• ccwhy session <id> (per-turn breakdown)
• ccwhy --json (machine-readable output)