The Agent Skills Directory

[SAFE]: No security threats were identified. The skill is designed to manage API costs locally.
[INDIRECT_PROMPT_INJECTION]: The skill reads session log files (~/.openclaw/sessions/*.jsonl) to estimate token usage. The ingestion process in check.py uses json.loads to extract numerical usage data for calculation. Because this data is used strictly for mathematical estimation and is not interpolated into agent prompts or shell commands, it does not present an injection vector.
[DATA_EXFILTRATION]: There are no network operations (such as curl, requests, or socket) present in the code. The skill only interacts with local files in the ~/.openclaw/ directory for reading logs and saving its own state.
[COMMAND_EXECUTION]: The Python script check.py does not utilize subprocess, os.system, or eval. It performs its logic using standard Python arithmetic and file I/O.
[OBFUSCATION]: The code is written in cleartext with no signs of encoding, zero-width characters, or hidden content.

spend-circuit-breaker