distribute-skill-to-all-agents

Warn

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell commands such as cp and rsync to move files between directories in the user's home folder. It interpolates the $SKILL variable directly into command strings without sanitization, creating a potential command injection vulnerability if the skill name is derived from untrusted input.\n- [PROMPT_INJECTION]: The skill enables the propagation of other skills across multiple agent environments (Claude, Codex, Pi, Hermes). This increases the attack surface for indirect prompt injection, as a malicious skill can be automatically synced to all agents on the system. Evidence Chain:\n
  • Ingestion points: Reads directory names and file paths from the user's home directory (e.g., ~/.agents/skills/).\n
  • Boundary markers: None. No explicit instructions are provided to the agent to ignore potentially malicious content within the skills being moved.\n
  • Capability inventory: Includes the ability to write to and overwrite files in multiple agent configuration directories (~/.claude/, ~/.pi/, ~/.hermes/).\n
  • Sanitization: No sanitization or validation is performed on the folder names or the contents of the SKILL.md files being distributed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 10, 2026, 02:46 PM
Security Audit — agent-trust-hub — distribute-skill-to-all-agents