image-loading

Pass

Audited by Gen Agent Trust Hub on Aug 2, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses Glob and Grep tools to search the user's project directory for existing image loading libraries. This is a standard pre-generation check to avoid library conflicts and is performed within the local project context.
  • [DATA_EXFILTRATION]: The generated Swift code includes an ImageDownloader that uses URLSession.shared.data(from: url) to fetch images from remote URLs provided at runtime. This is the intended primary purpose of the skill and does not involve accessing sensitive local system files or credentials.
  • [COMMAND_EXECUTION]: The DiskImageCache implementation uses FileManager to store images in the system's designated .cachesDirectory. It correctly generates filenames using SHA256 hashing of the URL strings, which effectively mitigates path traversal vulnerabilities by ensuring filenames do not contain directory navigation characters.
  • [PROMPT_INJECTION]: The SKILL.md instructions guide the agent through a logical flow of checking context, asking configuration questions, and generating code based on provided templates. No instructions attempting to bypass safety filters or override system prompts were found.
  • [REMOTE_CODE_EXECUTION]: The skill generates static Swift source code files. It does not download external scripts for execution or rely on unverified third-party package managers during the generation process.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 2, 2026, 01:18 PM
Security Audit — agent-trust-hub — image-loading