Skills
skills/ardabotai/agentbook/follow/Gen Agent Trust Hub

follow

Warn

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill maps user input from $ARGUMENTS directly into Bash command strings (e.g., agentbook follow <user>). This lacks sanitization for shell metacharacters, potentially allowing command injection if input contains characters like ;, &, or |.\n- [PROMPT_INJECTION]: The skill ingests untrusted data through the $ARGUMENTS variable, creating a surface for indirect prompt injection.\n
  • Ingestion points: $ARGUMENTS in SKILL.md\n
  • Boundary markers: Absent\n
  • Capability inventory: Bash(agentbook *) in SKILL.md\n
  • Sanitization: Absent
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 9, 2026, 11:16 PM