create-offer
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the execution of
setup.shandbuild.mjsto prepare its environment and perform the core task of generating PDF documents. - [EXTERNAL_DOWNLOADS]: The
setup.shscript downloads Node.js dependencies and the Chromium browser engine from the official npm registry and Playwright distribution servers to enable document rendering. - [DYNAMIC_EXECUTION]: The
build.mjsscript dynamically assembles HTML content at runtime from templates and JSON configuration data before rendering it via a headless Chromium browser instance. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied input to populate document templates, creating an attack surface for indirect instructions if processed data is malicious.
- Ingestion points: User input collected during the conversational interview described in
SKILL.mdand data read from the input JSON configuration file. - Boundary markers: The custom templating engine does not employ specific delimiters or "ignore instructions" warnings to isolate potentially untrusted content.
- Capability inventory: The skill utilizes file system write operations and controls a headless browser through the Playwright automation library.
- Sanitization: The template engine provides standard HTML escaping for typical substitutions but allows raw HTML injection via triple-brace syntax (
{{{...}}}) to support necessary rich-text formatting in the generated offers.
Audit Metadata