OpenRouter Free Helper

监控 OpenRouter 免费模型的到期通知和新模型发现，自动每日检查并推送飞书通知

This skill appears to do what it says, but before installing or enabling automated runs: 1) Inspect scripts/check-models.py (remaining portion not shown) to see exactly how notifications are sent (whether it reads tokens from ~/.openclaw/openclaw.json or calls an external CLI). 2) Verify the OpenClaw config (~/.openclaw/openclaw.json) contents so you understand which delivery credentials (Feishu tokens, channel configs) the skill might use. 3) Run the script manually in dry-run/--no-notify mode to confirm behavior and check the produced data/status.json. 4) Avoid copying the example cron payload with hardcoded /Users/neoshi paths — replace with paths appropriate to your account and confirm the job runs in an isolated session as intended. 5) If you are uncomfortable with automatic posting, keep cron-managed delivery enabled and invoke the script with --no-notify so the outer delivery mechanism (which you control) sends the final message.

Type: OpenClaw Skill Name: openrouter-free-helper Version: 1.0.9 The skill bundle is classified as suspicious because it performs high-risk operations, including programmatically launching Google Chrome with remote debugging enabled (`--remote-debugging-port=9222`) and reading the primary OpenClaw configuration file (`~/.openclaw/openclaw.json`), which typically contains sensitive API keys. While these actions are plausibly required for scraping dynamic content and integrating with the OpenClaw environment, they grant the script access to sensitive credentials and browser control. Additionally, `SKILL.md` and `cron-task.md` contain instructions that direct the AI agent to follow specific execution paths and suppress standard outputs, which is a form of prompt control that increases the attack surface of the agent.