The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check.sh uses the eval command to execute shell strings found in the check_cmd and latest_cmd fields of the references/tools.json file. This pattern allows for the execution of arbitrary commands defined within the skill's configuration.
[COMMAND_EXECUTION]: The skill is designed to facilitate the execution of arbitrary update_cmd shell strings (e.g., npm update, brew reinstall, pip3 install --upgrade) upon user confirmation. This provides a direct path for executing system-level changes through the agent.
[DATA_EXFILTRATION]: (System Reconnaissance) The scripts/discover.sh script performs a highly comprehensive scan of the user's local environment. It collects detailed lists of installed software from npm, Homebrew, pip, uv, conda, and searches binary directories like ~/.local/bin and /usr/local/bin. While this data is intended for the report, it provides the agent (and potentially any underlying LLM) with a complete profile of the user's installed software and environment configuration.
[COMMAND_EXECUTION]: The maintenance instructions in SKILL.md prompt the agent to execute a shell script located at ~/.openclaw/scripts/patch-health-timeout.sh. This involves executing code that resides outside of the skill's own package directory.

laohan-gengxin