French vocab automation. Formats Excel vocab to Google Sheet, generates ElevenLabs audio, uploads to Drive. Triggers: process french vocab, generate audio, F...

Before installing or running this skill: - Verify gog, gemini, and sag are the exact CLIs you expect and are installed from trusted sources. CLI supply-chain risk is real. - Confirm how your machine is authenticated to Google (gog). The skill will read and write Google Sheets and upload files to Drive; the repository does not declare the Google credentials/config required — ensure you only grant a least-privilege account or folder access, not broad account access. - Protect API keys: GEMINI_API_KEY and ELEVENLABS_API_KEY are required; treat them as secrets and avoid exposing them to shared environments. - Review and harden scripts: replace the shell=True pattern in scripts/fix_update.py (which injects /tmp/french_learning_output_data.json via a shell substitution) with a safe non-shell subprocess call that reads the file directly or passes JSON as an argument. The current pattern could allow command injection if an attacker can control the temp file. - Consider running first in a sandbox or with a restricted Google service account that only has write access to the intended Sheet/Drive folder, and audit the folder after a test run. - If you do not want the skill to upload to your Drive, do not provide the Drive Folder ID or run the generate_audio script until you have validated outputs locally. If you want, I can point out the exact lines to change to eliminate the shell invocation and make the scripts safer.

Type: OpenClaw Skill Name: french-learning Version: 1.0.0 The skill contains a shell injection vulnerability in `scripts/fix_update.py`, which uses `subprocess.run(shell=True)` combined with shell command substitution `$(cat ...)` to process a temporary JSON file. While the skill's stated purpose of automating French vocabulary via Google Sheets and ElevenLabs (using the `gog` and `sag` tools) appears legitimate, this coding practice allows for potential arbitrary command execution if the contents of the temporary file are manipulated. No clear evidence of intentional malice or data exfiltration was found, but the inclusion of an insecure update script warrants a suspicious classification.