wechat-2d-render
Warn
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill clones a repository from a third-party GitHub user (sxhzju/wechat-2d) at runtime.- [REMOTE_CODE_EXECUTION]: The script executes 'pnpm install' and 'pnpm run' within the cloned repository. This process downloads and executes arbitrary code or build scripts defined in the external package.json, which are not under the control of the skill's author.- [COMMAND_EXECUTION]: The script runs shell commands using workspace paths and file locations provided by the user. While the script uses double quotes to prevent basic shell splitting, it remains a surface for path-related issues if the agent provides unsanitized input.- [INDIRECT_PROMPT_INJECTION]: The skill ingests a user-provided JSON props file, which constitutes a surface for indirect instructions if the agent or the underlying scripts process this data to make decisions.
- Ingestion points: Custom props_file path parameter in 'scripts/render_wechat_2d.sh'.
- Boundary markers: Absent; the content is passed directly to the render engine.
- Capability inventory: File system access, network access via git and pnpm, and arbitrary script execution via pnpm run.
- Sanitization: None provided for the content of the ingested file.
Audit Metadata