cli-skill-collector

Warn

Audited by Gen Agent Trust Hub on Jul 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the global installation of the omniroute package via npm and facilitates downloading skill files from arbitrary GitHub repositories.
  • [COMMAND_EXECUTION]: The omniroute tool includes subcommands like setup, update, and env set, which allow for modification of system configurations and environment variables.
  • [REMOTE_CODE_EXECUTION]: The installation workflow directs the agent to download SKILL.md files from GitHub and copy them into executable context directories such as ~/.claude/skills/. This allows for the introduction of external code or instructions that could execute in the agent's environment.
  • [PROMPT_INJECTION]: By installing external skills from GitHub into local agent directories, the skill creates a surface for indirect prompt injection. Malicious instructions embedded in a downloaded repository could hijack the agent's behavior once the skill is active.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 16, 2026, 07:56 AM
Security Audit — agent-trust-hub — cli-skill-collector