cursor-cli

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches installation script from the official Cursor website (cursor.sh).
  • [REMOTE_CODE_EXECUTION]: Executes the installation script by piping the remote content directly to the shell.
  • [PROMPT_INJECTION]: The skill documents how to pass untrusted natural language data through CLI flags like -p.
  • Ingestion points: Prompt interpolation into command line arguments as documented in SKILL.md.
  • Boundary markers: Absent from documentation examples.
  • Capability inventory: The agent command can edit files, generate code, and communicate with remote cloud environments.
  • Sanitization: No validation or sanitization of input prompts is documented.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 06:45 PM
Security Audit — agent-trust-hub — cursor-cli