sites-building

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains instructions that prioritize autonomy and task completion over typical interaction boundaries, including directives to "never self-limit," "never pause to ask discovery questions," and "Proceed immediately with best judgment." These patterns are designed to prevent the agent from stopping during complex tasks but mirror techniques used to bypass safety constraints.
  • [COMMAND_EXECUTION]: The skill is designed to execute shell commands within the environment to manage the site lifecycle. This includes running a local initialization script (scripts/init-site.sh), performing port scanning for server diagnosis, and executing Node.js build and development scripts (npm run dev, npm run build).
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests and processes untrusted data from user requests and web image searches to generate site source code and metadata.
  • Ingestion points: User-supplied site descriptions and content; data returned from web image searches.
  • Boundary markers: None identified in the instructions to separate untrusted external data from system instructions.
  • Capability inventory: Local shell execution (scripts/init-site.sh, npm), browser-based previewing (open_in_codex), and deployment to hosting services (sites-hosting).
  • Sanitization: There are no explicit instructions for sanitizing or validating external content before it is interpolated into the site's code or social preview metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 12:17 AM