hiapi-seedream-5-0-pro
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The installer script retrieves the skill's source code by cloning the vendor's repository from GitHub during the setup and update process.
- [COMMAND_EXECUTION]: The installation logic executes
gitcommands to facilitate skill deployment and management within the agent's local environment. - [EXTERNAL_DOWNLOADS]: The skill fetches update manifests from the vendor's official GitHub repository to ensure version compatibility and downloads generated image files from the HiAPI platform for local storage.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and external image URLs to generate content, which constitutes a surface for handling untrusted data.
- Ingestion points: User-provided prompt text and reference image URLs via the
input-urlparameter inscripts/hiapi-seedream-5-pro.mjs. - Boundary markers: Prompt content is processed as raw text; reference URLs are delimited and validated for protocol and supported image formats.
- Capability inventory: The skill possesses file system write access in
saveImageOutputsand network request capabilities for API interaction. - Sanitization: The
normalizeInputUrlsfunction inscripts/lib/seedream-5-pro.mjsvalidates that URLs use the HTTP/HTTPS protocols and explicitly rejects unsupported file types such as SVG to prevent potential parsing exploits.
Audit Metadata