prototype
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze "surrounding code" to determine prototyping goals. This creates a surface where an attacker with control over the codebase could inject malicious instructions that the agent might inadvertently execute while generating the prototype. \n- Ingestion points: Surrounding code and user prompts (SKILL.md). \n- Boundary markers: None specified for code analysis. \n- Capability inventory: File writing, git operations, and execution of task runner commands like
pnpm,python, andbun(SKILL.md). \n- Sanitization: No sanitization of project context is performed.\n- [COMMAND_EXECUTION]: The skill encourages the agent to provide or use shell commands (e.g.,pnpm,python,bun) to execute prototypes. While standard for development, this capability could be exploited if malicious instructions are injected via the project context.
Audit Metadata