new-workspace

Fail

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DATA_EXFILTRATION]: The skill is instructed to read ~/.claude/CLAUDE.md to extract 'user identity facts', OS details, locale, and timezone information.
  • [DATA_EXFILTRATION]: By default, the skill creates a public GitHub repository using gh repo create --public and pushes the scaffolded workspace, which contains the harvested identity facts and ambient metadata, to a public remote server.
  • [COMMAND_EXECUTION]: User-supplied positional arguments from $ARGUMENTS (such as the workspace name) are used directly in shell commands including mkdir, git commit, and gh repo create. The instructions do not specify any sanitization or validation of these inputs, which could allow for command injection if a malicious workspace name containing shell metacharacters is provided.
  • [EXTERNAL_DOWNLOADS]: The skill performs network operations using the GitHub CLI (gh) and git to create repositories and push data to GitHub's servers.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 17, 2026, 04:31 PM
Security Audit — agent-trust-hub — new-workspace