packaging-workshop

Warn

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes instructions to execute a shell command: python3 ~/.grok/skills/zan-punct-fix/scripts/fix_punctuation.py <目标路径> --dry-run. This pattern directly interpolates a user-controlled variable (<目标路径>) into a shell command. Without proper sanitization by the underlying agent platform, this could allow an attacker to execute arbitrary commands by providing a malicious path containing shell metacharacters (e.g., ;, &&, |).
  • [EXTERNAL_DOWNLOADS]: The skill relies on an external dependency named zan-punct-fix located at a specific local path (~/.grok/skills/zan-punct-fix/). While the skill does not download the dependency at runtime, it creates a hard dependency on external code that is not contained within the skill itself, increasing the attack surface if that external skill is compromised or manipulated.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 21, 2026, 05:00 PM
Security Audit — agent-trust-hub — packaging-workshop