skills/ferueda/harness/cursor-cli/Gen Agent Trust Hub

cursor-cli

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the official Cursor CLI using a piped-to-bash command (curl https://cursor.com/install | bash) in lib/home.ts and SKILL.md. Since Cursor is a well-known developer tool, this reference to the official installation source is documented neutrally.
  • [COMMAND_EXECUTION]: The skill executes the local agent CLI binary via the Node.js spawn API in lib/runner.ts. It constructs commands based on user input, passing flags and prompts to the underlying Cursor agent. The execution is limited to the agent executable resolved from the system PATH or standard installation locations.
  • [PROMPT_INJECTION]: The skill provides an interface for delegating tasks, creating a surface for indirect prompt injection. \n
  • Ingestion points: User input is ingested in scripts/cursor-cli.ts through command-line arguments, file reading (--prompt-file), or stdin (--stdin). \n
  • Boundary markers: While lib/schema.ts wraps prompts in instructional headers for structured output, there are no specific boundary delimiters separating the user input from the system instructions. \n
  • Capability inventory: The skill has the capability to execute the Cursor sub-agent via lib/runner.ts, which may perform file system or network operations based on the prompt instructions. \n
  • Sanitization: The skill does not perform sanitization on the input strings before interpolating them into the sub-agent command, which is typical for a wrapper of this type.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 01:45 AM
Security Audit — agent-trust-hub — cursor-cli