skill-reflect

Warn

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill exhibits a command injection vulnerability where user-provided inputs for <skill-name> and <plugin> are directly interpolated into shell commands. An attacker could provide a malicious string containing shell metacharacters (e.g., ;, &&, |, $(...), or backticks) to execute arbitrary commands with the user's permissions.
  • [COMMAND_EXECUTION]: Vulnerable command patterns identified in the skill instructions include:
  • ls -d plugins/*/skills/<skill-name>/ 2>/dev/null (used in Phase 1 to find the plugin)
  • mkdir -p ~/.claude/skill-sessions/<plugin>/sessions/<skill-name> (used in Phase 3 to create the log directory)
  • ~/.claude/skill-sessions/<plugin>/sessions/<skill-name>/<timestamp>.md (used in Phase 3 as the destination path for log creation)
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 25, 2026, 02:28 PM
Security Audit — agent-trust-hub — skill-reflect