design

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. User-supplied inputs such as brand names and design prompts are interpolated directly into AI model templates in scripts like scripts/logo/generate.py, scripts/cip/generate.py, and scripts/icon/generate.py without sanitization or boundary markers. This could allow an attacker to provide a malicious brand name that overrides the agent's instructions.
  • [COMMAND_EXECUTION]: Documentation in references/social-photos-design.md provides examples for running shell commands and automation scripts (e.g., Chrome headless, Playwright) that use variables for dimensions and file paths. If these are implemented by the agent using untrusted input without proper validation, it could lead to command injection vulnerabilities.
  • [DATA_EXFILTRATION]: The generation scripts are configured to read .env configuration files from the user's home directory (~/.claude/.env). While used for API key management, accessing credentials from global home directory paths is a sensitive operation that should be monitored for unintended data access.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 03:44 PM
Security Audit — agent-trust-hub — design