The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages system configurations through local shell scripts (backup_config.sh, restore_config.sh, etc.) that interact with the openclaw CLI tool. These operations are restricted to the local environment and the application's specific configuration directory.- [PROMPT_INJECTION]: Guardrails defined in SKILL.md explicitly instruct the agent to seek user approval before restarting services or applying changes, effectively mitigating the risk of unauthorized or accidental system modifications.- [SAFE]: The validate_config.sh script implements defensive measures by checking for the presence of dangerous keywords like 'sudo', 'exec', or 'shell' within the configuration data. It also performs consistency checks on model configurations using embedded Python logic, ensuring the integrity of the setup before deployment.

config-guardian