Openclaw Whisperer

Comprehensive diagnostic, error-fixing, and skill recommendation tool for OpenClaw

This repository looks like a legitimate diagnostic/repair tool, but several red flags mean you should not run it blindly: 1) SECURITY.md claims 'no network calls', yet fix-recipes and references include docker pull, wget (GitHub release), npm installs and openclaw skill installs — ask the author to clarify and provide an explicit network policy. 2) The package does not declare any required environment variables, but many recipes reference sensitive tokens (OPENAI_API_KEY, GATEWAY_TOKEN, Slack/Discord tokens). Verify exactly when/where the tool will read or transmit secrets and avoid providing credentials until reviewed. 3) Several fix recipes run destructive system commands (kill -9, systemctl start, cp/mv, chmod) and shell pipelines (lsof | xargs kill) — inspect how the Python code executes recipe commands (does it use shell=True?) and test in a safe environment first. 4) The self-updater and runtime downloads increase risk; prefer running the code in an isolated/sandboxed environment or container, review/upstream hashes for downloaded artifacts, and disable automatic updates until you trust the source. If you plan to use auto-fix features, back up your openclaw.json and relevant data first and enable interactive confirmation for risky recipes. If possible, request the maintainer to: (a) document exact network endpoints used, (b) declare required env vars in the manifest, and (c) remove or explicitly mark any remote-download steps so you can vet them before execution.

Type: OpenClaw Skill Name: openclaw-whisperer Version: 1.2.0 The skill bundle is classified as suspicious due to high-risk capabilities, specifically the execution of `npm install -g` commands without explicit user confirmation when the `--auto-fix` flag is used. While the stated purpose is to install core OpenClaw components (`pnpm`, `openclaw` CLI), this action (defined as `safe_auto: true` in `data/fix-recipes.json`) downloads and executes code from external registries, posing a significant supply chain risk. The `SECURITY.md` documentation's examples for `safe_auto: true` (info messages, config reads, version checks) do not align with the actual execution of software installations, indicating a potential misrepresentation of safety. Other powerful commands like `kill -9`, `chmod`, and `docker rm` are appropriately marked `safe_auto: false` (requiring confirmation), but the breadth of system-level operations still presents a high-risk profile for a diagnostic tool.