Skills
Navigate
skills/msapps-mobile/claude-plugins/daily-cowork-backup/Gen Agent Trust Hub

daily-cowork-backup

Fail

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill programmatically reads a local configuration file, cowork-gdrive-config.json, to extract a sensitive apiKey. This secret is subsequently transmitted over the network.
  • [DATA_EXFILTRATION]: The skill archives sensitive directories—including the agent's internal skills store, remote plugins, and the configuration folder containing the apiKey—into zip files. These archives are then transmitted to a remote URL via an automated POST request.
  • [REMOTE_CODE_EXECUTION]: The skill dynamically generates a Python script (do_upload.py) on the host machine's filesystem and uses the Desktop Commander tool to execute it natively. This crosses the execution boundary between the agent's sandbox and the host operating system.
  • [COMMAND_EXECUTION]: The skill utilizes mcp__Desktop_Commander__start_process to execute high-risk operations on the host, including a recursive find ... -delete command to remove image files from the user's documents folder and the execution of the locally generated upload script.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 28, 2026, 01:39 PM
Security Audit — agent-trust-hub — daily-cowork-backup