create-brand-kit
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill interfaces with Google Chrome in headless mode to render brand assets. This is performed via
subprocess.run(inscripts/optics.py) andexecFileSync(inscripts/context.mjs,scripts/preview.mjs, andscripts/sheet.mjs). These calls use argument lists rather than shell strings, mitigating command injection risks. - [REMOTE_CODE_EXECUTION]: The skill features an optional integration with the
Codex CLIusing thecodex execcommand for generating AI mockups. This feature is gated by a check for the CLI's presence and requires explicit user consent. - [EXTERNAL_DOWNLOADS]: The skill retrieves external data using
WebFetchandWebSearchfor brand research and downloads font files from Google Fonts. These are standard operations for the skill's purpose. - [PROMPT_INJECTION]: The skill documentation (in
SKILL.md) includes specific instructions to handle untrusted third-party content. It directs the agent to mine fetched pages for facts while explicitly ignoring any embedded instructions or commands, effectively mitigating indirect prompt injection attempts from malicious websites.
Audit Metadata