pc-assistant

PC healthcheck and diagnostics with detailed system information and actionable recommendations. Works on Windows, macOS, and Linux. Read-only system diagnost...

This appears to be a coherent local healthcheck skill, but it gathers sensitive system information (auth logs, SSH authorized_keys content snippets, package lists, mounted Windows user directories under WSL, etc.). Before installing or running: 1) Review the scripts yourself (they're included) and confirm you are comfortable with the files they read and where they write reports. 2) Run as a regular (non-root) user — the scripts are intended to be read-only but running as root increases exposure. 3) Choose a private output directory (do not leave reports in world-readable locations) and enable cleanup if desired. 4) If you schedule it, add cron entries yourself rather than granting automatic scheduling. 5) If you require the skill to be invoked from a different path than ~/.npm-global/..., adjust the call paths or run scripts directly. If you want additional assurance, test in a VM or non-production machine first.

Type: OpenClaw Skill Name: pc-assistant Version: 1.2.5 The skill performs an extremely broad system diagnostic that collects highly sensitive information, including environment variables (via `env`), shell history (`.bash_history`), and SSH `authorized_keys` in `scripts/healthcheck.sh`. While the tool is transparently documented as a healthcheck utility and includes warnings in `install.yaml` and `SKILL.md` regarding the sensitivity of the generated reports, the collection of shell history and environment variables poses a significant risk of leaking API keys, passwords, or other secrets to the AI agent. There is no evidence of intentional exfiltration to a remote server, but the data collection is overly invasive for its stated purpose.