Real-time AI API usage tracking and cost monitoring for OpenClaw. Track spending across OpenAI, Claude, Gemini, Kimi, DeepSeek, and Grok with live dashboard....

Do not run the included installer or setup scripts without manual review. Specific risks to consider: - The collector reads agent session files (/root/.openclaw/agents/...) and the OpenClaw config — these can contain sensitive runtime/session metadata; confirm you are comfortable with that access. - Installer.sh uses sudo to install systemd services and writes to system locations (/etc/systemd/system, /var/www/html). That gives persistent, privileged background processes; only allow if you trust the code and run it on an isolated host. - Several repository and path names are inconsistent (bill-project vs project-bill vs ai-bill etc.) and SKILL.md paths don't always match code. This could be sloppy or indicate copy-paste errors; it is a red flag — verify the canonical source repository and inspect it yourself. - The web UI includes weak client-side locking (hard-coded SECURE_CODE = "1234") and the collector writes usage to /var/www/html, which may expose billing/session data publicly without authentication. Recommendations: 1) Inspect the full code locally (collector.js, installer.sh, setup.js) before running anything. Search for all file reads/writes and network calls. 2) If you want to try it, run on an isolated VM/container (not your primary machine) and do not run installer.sh with sudo until you’ve audited the service unit files it would install. 3) Remove or change any code that writes usage to a public webroot, and protect the dashboard with a real server-side auth mechanism. 4) Confirm the authoritative GitHub repository and verify checksums/releases instead of piping curl|bash from an unknown URL. 5) If you cannot audit the code, prefer an alternative tool from a known, trusted source.

Type: OpenClaw Skill Name: ai-bill-clawhub Version: 2.2.5 The skill is classified as suspicious due to several high-risk capabilities and questionable practices. The `app/collector.js` script reads sensitive OpenClaw agent session data (`/root/.openclaw/agents/main/sessions/sessions.json` and `/root/.openclaw/openclaw.json`) and performs high-privilege writes to system-wide web server directories (`/var/www/html/bill/usage_live.json`, `/var/www/html/bill/usage.json`). Additionally, the `app/sync_prices.sh` script declares fetching prices from a specific GitHub repository (`openclaw/project-bill`) but then hardcodes the content, creating a discrepancy that raises supply chain concerns. A hardcoded weak password ('1234') for a client-side lock screen is also present in `app/index.html`, indicating poor security hygiene. While these actions are not explicitly malicious, they represent significant vulnerabilities and powerful capabilities that could be exploited.