The Agent Skills Directory

[COMMAND_EXECUTION]: The skill features a wide array of bash commands for monitoring processes, inspecting logs, and querying SQLite databases. These commands are intended for system administration and status checks.
[CREDENTIALS_UNSAFE]: Diagnostics include reading configuration files like openclaw.json and session logs. Although the text advises redacting sensitive info, the provided jq and cat commands do not include mechanisms to mask credentials such as botToken or apiKey, potentially exposing them in the terminal or to the agent context.
[EXTERNAL_DOWNLOADS]: The documentation encourages the use of clawdhub and npx to install additional skills and plugins. This facilitates the introduction and execution of third-party code from external repositories into the local environment.
[REMOTE_CODE_EXECUTION]: The orchestration section describes how to spawn background agents (e.g., Codex or Claude) using the bash tool with high-privilege flags like --yolo and pty:true, enabling autonomous execution of tasks and code without immediate human oversight.

openclaw-config