The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard shell commands like cat, mkdir, echo, and jq to create and query local tracking files in the ~/.claude/ directory. These commands are used as intended for managing session state and generating cost reports.
[DATA_EXPOSURE]: The skill reads and writes to specific files in the ~/.claude/ directory to track API spend. It does not attempt to access sensitive system files, environment variables containing secrets, or user credentials.
[PROMPT_INJECTION]: Instructions regarding model routing (e.g., using specific models for specific task types) are functional guidelines for cost optimization and do not contain attempts to override core safety filters or agent behavior.
[SAFE]: The skill operates entirely locally using the agent's native command capabilities for its stated purpose of cost metering. No external network requests or remote code execution patterns were detected.

cost-metering