compose-preview-design-board
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The documentation in
SKILL.mdincludes an installation command that fetches a script fromhttps://raw.githubusercontent.com/yschimke/skills/main/scripts/install.shand pipes it directly tobash. This pattern is used to install the required CLI tools and is directed at the author's own repository. - [COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands, specifically
compose-previewto render images andpython3 scripts/build-design-board.pyto assemble the final design board artifact. - [INDIRECT_PROMPT_INJECTION]: The Python script
scripts/build-design-board.pyprocesses external JSON specifications which could potentially contain malicious instructions. The script follows security best practices by usinghtml.escape()to sanitize all text fields (captions, notes, titles, etc.) before embedding them into the generated HTML file, preventing potential XSS or injection when the board is viewed.
Audit Metadata