openclaw-config-helper
Fail
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill's documentation (Case 3 in
SKILL.md) explicitly instructs the AI that API keys must be hardcoded in the configuration file, stating '必须硬编码真实 key' (must hardcode real key). This promotes the storage of sensitive secrets in plain text, which is a major security risk. - [DATA_EXFILTRATION]: The skill accesses and retrieves the content of
~/.openclaw/openclaw.jsonthrough thegateway action=config.getcommand and thescripts/check_config.shscript. This configuration file typically contains sensitive authentication tokens, account IDs, and private settings for messaging channels like Telegram and WhatsApp. - [COMMAND_EXECUTION]: The skill frequently executes system commands using the
gatewaytool to modify the system state (config.patch,config.apply). Additionally, it references and invokes external scripts from other skills, such as~/clawd/skills/tavily/scripts/tavily.sh, which represents an external dependency and command execution surface. - [EXTERNAL_DOWNLOADS]: The skill uses
web_fetchto retrieve configuration guidelines fromdocs.openclaw.ai. This is documented as a neutral action intended for retrieving official project documentation.
Recommendations
- AI detected serious security threats
Audit Metadata