OSV Scanner

Scans Node.js/NPM and Linux packages for known vulnerabilities using the Google OSV API by analyzing dependencies and reporting related CVEs.

This skill is coherent for vulnerability scanning and doesn't ask for secrets, but review and fix before running: 1) scanner.py has a clear bug in the npm branch: the f-string lines.append(f'{name}=={version.split('^')[1]}') uses conflicting quotes and will cause a syntax error (and will break for versions without '^'). Fix by stripping '^' or using version.lstrip('^') and proper quoting. 2) SKILL.md asks you to rename template/ScanResult.txt → template/ScanResult.xlsx because the script expects the .xlsx path; do that or change TEMPLATE_PATH. 3) The script uses subprocess(shell=True) for static commands (lsb_release, dpkg-query) — these are reasonable for enumerating OS packages but avoid running the script on systems with sensitive data or where you do not trust included files. 4) The script POSTs package info to api.osv.dev (expected); if you need to avoid sending data externally, do not run it or modify it to use a local DB. 5) As always, run untrusted scripts in an isolated environment (container or VM), inspect/patch the code (npm branch and error handling) before use, and verify the template/header ('MachineA') matches your expected spreadsheet format.

Type: OpenClaw Skill Name: osv-scanner Version: 1.0.0 The skill bundle is classified as suspicious due to two main indicators. First, the `SKILL.md` and `README.md` files contain an explicit instruction to rename `template/ScanResult.txt` to `template/ScanResult.xlsx`. The `ScanResult.txt` file is actually a ZIP archive (an XLSX file disguised with a .txt extension), which is a suspicious file manipulation pattern that could be used for evasion. Second, the `scanner.py` script utilizes `subprocess.run` with `shell=True` for executing system commands (`lsb_release`, `dpkg-query`). While the commands are hardcoded and not directly exploitable in this specific context, the use of `shell=True` is a risky capability that generally increases the attack surface for shell injection vulnerabilities if any part of the command string were to become user-controlled or untrusted.