upscale-image
Pass
Audited by Gen Agent Trust Hub on May 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references upscayl.org for manual installation of dependencies. This is an informational link and does not involve automated downloads or remote code execution by the agent.
- [COMMAND_EXECUTION]: The skill's procedure involves executing shell commands (upscayl-bin, vipsthumbnail, cjpeg-mozjpeg, and rm) using variables derived from user input, such as file paths and model names. While the templates use double quotes, they lack instructions for sanitizing these inputs, which could be exploited for command injection if filenames contain shell-special characters.
- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted data from the file system.
- Ingestion points: File paths and directory paths provided as input to the agent (SKILL.md, Procedure section).
- Boundary markers: Absent. There are no instructions to the agent to treat these inputs as literal strings or to ignore instructions embedded in metadata or filenames.
- Capability inventory: Execution of several binaries via shell and file deletion capabilities.
- Sanitization: Absent. The procedure does not specify validation or escaping logic for the inputs before they are passed to the shell environment.
Audit Metadata