parallel-design-variants
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use the command
python3 -m http.serverto serve the generated design gallery locally for review. This is a standard development practice for previewing static HTML files. - [INDIRECT_PROMPT_INJECTION]: The workflow involves ingesting 'real data' and 'constant content' provided by the user or external sources to populate the design variants. This represents a potential attack surface for indirect prompt injection if the input data contains malicious instructions intended to influence the subagents' behavior.
- Ingestion points: Untrusted content enters the workflow via the 'constant content' section of the orchestration document (spec issue) described in
SKILL.md. - Boundary markers: No specific delimiters or 'ignore' instructions are provided to separate user-supplied content from the subagent instructions.
- Capability inventory: The skill utilizes parallel subagent tool calls, file system writes to create HTML/JS variants, and shell command execution for the local preview server.
- Sanitization: There is no explicit requirement for the agent to sanitize or escape the input data before interpolating it into the subagent scaffolds.
Audit Metadata