advai-cli-unified-terminal-ai

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install the advai-cli package via PyPI (pip install advai-cli) and npm (npm install -g advai-cli). These are vendor-owned packages associated with the skill author 'Aradotso'.
  • [REMOTE_CODE_EXECUTION]: The advai skill install and advai cli install commands allow downloading and installing content directly from GitHub repositories. While the examples use placeholders or well-known tools like the GitHub CLI, this capability involves the fetching and syncing of remote code or configuration into the agent's environment.
  • [COMMAND_EXECUTION]: Python code examples provided in the documentation (setup_advai, deploy_skill) utilize subprocess.run to execute shell commands for package installation and tool configuration.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection via its knowledge base and skill installation features.
  • Ingestion points: Untrusted data enters the context through advai kb doc add (which ingests local documents) and advai skill install (which imports skill definitions from external GitHub URLs).
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are documented for these ingestion paths.
  • Capability inventory: The tool possesses capabilities to execute subprocesses, install packages, and write files to sensitive agent platform directories (e.g., ~/.cursor/skills).
  • Sanitization: There is no evidence of sanitization, filtering, or validation for the content of ingested documents or external skill definitions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 06:34 PM
Security Audit — agent-trust-hub — advai-cli-unified-terminal-ai