The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts (fetch_usage.py, preflight_check.py, subagent_guard.py) use the subprocess module to execute system commands. This is primarily used to interact with the openclaw CLI tool for browser automation and to chain execution between internal scripts. Analysis shows that arguments passed to these commands, such as target_id, are strictly validated using regular expressions (e.g., hexadecimal patterns), which effectively prevents shell command injection.
[EXTERNAL_DOWNLOADS]: The skill requires the openclaw CLI tool and a Chrome extension to function. These are external dependencies that the user must provide in their environment. The scripts themselves do not perform any hidden or unprompted downloads.
[PROMPT_INJECTION]: The SKILL.md file includes instructions that guide the agent's autonomous behavior based on current usage levels (e.g., 'Disable non-essential reasoning' when capacity is low). These are functional heuristics for the agent's self-management and do not attempt to bypass core safety filters or override the system prompt in a malicious manner.
[DATA_EXPOSURE_AND_EXFILTRATION]: While the skill accesses browser tab data via the openclaw tool, it only processes this data locally to extract usage percentages and timestamps. There is no evidence of data being transmitted to external or unauthorized domains.

kimi-usage-monitor