OpenClaw Power Ops

Operate and maintain OpenClaw installations — CLI commands, config management, channel/agent/model setup, security auditing, troubleshooting, and gateway adm...

This skill appears to be a legitimate OpenClaw admin reference, but proceed cautiously because the source is unknown and some instructions can expose or change secrets. Before installing or invoking: 1) Verify the skill content against official OpenClaw docs at https://docs.openclaw.ai; 2) Back up openclaw.json, credentials/, and agent workspaces before running --fix commands; 3) Avoid running any LLM audit that sends your ~/.openclaw or credential files to a remote service — if you must use an LLM, ensure it's configured to run locally or redact secrets first; 4) Prefer running audit/doctor commands in read-only or dry-run mode first and review proposed fixes manually; 5) Confirm any generated tokens/permission changes (e.g., gateway.auth.token, chmod changes) are applied intentionally. If you don't trust the unknown owner, treat this as documentation only and perform changes manually after review.

Type: OpenClaw Skill Name: openclaw-power-ops Version: 1.0.0 The skill bundle provides extensive administrative capabilities, including the execution of `openclaw` CLI commands for system configuration, agent/channel management, and gateway control, as well as direct shell commands like `chmod` and `openssl rand` (in `references/security-audit.md`). While these capabilities are necessary for the stated purpose of 'OpenClaw Operations' and are used for legitimate tasks like security hardening (e.g., generating tokens, fixing permissions), they represent a significant attack surface. The `SKILL.md` also contains prompt injection instructions for the agent to perform a 'Deep Audit with Claude Code', demonstrating the agent's ability to interpret and act on such directives. The broad access and command execution capabilities, even if intended for benign administration, elevate the risk to 'suspicious' due to the potential for misuse if the agent were to be compromised by a malicious user prompt.