openclaw-config-helper

Warn

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill explicitly advises users to hardcode plain-text API keys (e.g., OpenAI 'sk-' keys) into the configuration files. In 'Case 3' of the SKILL.md file, it specifically recommends against using secure credential references (like pass:), stating that they are not supported and that the user 'must hardcode the real key'. This practice leads to the exposure of sensitive secrets in plain-text files.
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute the gateway command-line utility to modify system configurations (config.patch and config.apply). Although it includes a step to wait for user confirmation, this capability allows for the programmatic alteration of the application's core settings. Additionally, the helper script scripts/check_config.sh reads the contents of ~/.openclaw/openclaw.json using jq. This configuration file typically contains sensitive information, such as communication channel tokens for Telegram and WhatsApp, which are then output to the console and exposed to the agent's context.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 17, 2026, 10:06 AM