install-renku

Warn

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute shell scripts directly from the vendor's infrastructure using curl | sh on macOS and irm | iex on Windows. These piped-to-shell patterns allow for arbitrary remote code execution on the host machine.\n- [DYNAMIC_EXECUTION]: The skill performs automated launch of the Renku Studio server using dynamic script generation and execution at runtime. On macOS, it generates and runs an AppleScript (osascript) to open a terminal and execute the start command. On Windows, it uses a Base64-encoded PowerShell command string via the -EncodedCommand flag to initiate the server in a new shell instance.\n- [COMMAND_EXECUTION]: Extensive local command execution is performed for environment discovery and runtime management, including operating system and architecture identification (uname, OSVersion.Platform), and direct interaction with the renku CLI tool for status checks and initialization.\n- [EXTERNAL_DOWNLOADS]: The skill fetches installation resources from https://downloads.gorenku.com. While these are vendor-provided assets for the skill's primary purpose, the delivery mechanism bypasses standard package management and verification workflows.\n- [INDIRECT_PROMPT_INJECTION]: The skill instructions create an attack surface for indirect prompt injection by directing the agent to automate and monitor a browser-based user interface which may process external data.\n
  • Ingestion points: Untrusted content could enter the agent context through the inspection and automation of the Renku Studio configuration and settings pages (found in SKILL.md).\n
  • Boundary markers: The instructions include specific guidance for the agent to avoid capturing sensitive API keys from the UI, directing users to input secrets directly into the Studio application instead of the chat context.\n
  • Capability inventory: The skill has capabilities for shell execution (CLI tools, installation scripts), automated script generation (AppleScript, PowerShell), and network access for downloading components.\n
  • Sanitization: The provided automation examples utilize quoted form in AppleScript and string escaping in PowerShell to reduce the risk of command injection during the assembly of launch commands.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 21, 2026, 04:06 PM