Skills from clawhub.ai/dgriffin831

This skill enables Claude Code Chrome browser automation via `nodes.run` and `claude` with a specified install path ` /opt/homebrew/bin/claude`. It instructs running shell commands (`nodes.run` examples), allows arbitrary web requests (e.g., `example.com`), and urges `--dangerously-skip-permissions`, creating elevated execution and network risk.

`skill-scan` is a security scanner for OpenClaw agent skills that performs static analysis and optional LLM deep inspection. It reads `OPENAI_API_KEY`/`ANTHROPIC_API_KEY`, posts scanned files to `https://api.openai.com/v1/chat/completions`/`https://api.anthropic.com/v1/messages`, runs `clawhub`/`openclaw` CLI commands, and instructs adding templates to `AGENTS.md`.

Input Guard scans untrusted external text for prompt-injection and optionally uses LLMs/taxonomy data. It performs network calls to `https://api.openai.com/v1/chat/completions`, `https://api.anthropic.com/v1/messages`, and `https://api.promptintel.novahunting.ai/api/v1`, uses `OPENAI_API_KEY`/`ANTHROPIC_API_KEY`/`PROMPTINTEL_API_KEY`, runs local commands (`openclaw`, `python3` scripts) and instructs adding a section to `AGENTS.md`.