happy-image-gen
Pass
Audited by Gen Agent Trust Hub on Apr 17, 2026
Risk Level: SAFE
Full Analysis
- [CREDENTIALS_UNSAFE]: No hardcoded credentials were found. The skill uses environment variables and supports 1Password integration for API key management, following best practices for CLI tools.\n- [DATA_EXFILTRATION]: The skill reads local file content (prompts and reference images) and sends it to external AI service providers such as OpenAI and Google. This behavior is consistent with the skill's stated purpose of generating images from local inputs, and all destination domains are well-known services.\n- [COMMAND_EXECUTION]: The setup documentation suggests standard installation commands for the Bun runtime from its official source (bun.sh). No unauthorized or hidden command execution patterns were found in the functional code.\n- [REMOTE_CODE_EXECUTION]: Dynamic code loading is used to import provider modules, but it is restricted to a predefined list of valid provider names, preventing arbitrary local file inclusion or execution.\n- [PROMPT_INJECTION]: The skill processes untrusted local data from files passed via CLI arguments. While this is an intended feature, it constitutes an indirect prompt injection surface.\n
- Ingestion points: scripts/main.ts reads prompt files; provider scripts read reference images.\n
- Boundary markers: Absent; files are concatenated with simple newlines.\n
- Capability inventory: Network access via fetch in all provider scripts to send data to external APIs.\n
- Sanitization: Absent; file contents are used directly in API payloads.
Audit Metadata