wizard
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed to help developers create ephemeral setup wizards. No malicious patterns, obfuscation, or unauthorized exfiltration attempts were found.- [DATA_EXFILTRATION]: While the skill handles secrets, it captures them through interactive user prompts and stores them using standard tools like .env files and the GitHub CLI (gh secret set). It does not send data to any external third-party servers.- [COMMAND_EXECUTION]: The generated bash scripts execute local commands to open URLs and manage repository settings. These actions are performed transparently as part of an interactive session driven by the user.- [PROMPT_INJECTION]: The skill analyzes project files to determine setup requirements, which is a surface for indirect prompt injection. However, the process includes a mandatory human-in-the-loop confirmation step before the wizard script is generated, mitigating risks from malicious content in repository files.
- Ingestion points: Processes repository files like README, .env.example, and workflow files (SKILL.md Step 1).
- Boundary markers: The instructions do not define specific delimiters for untrusted project content.
- Capability inventory: Includes file writing (write_env), secret management (gh secret set), and opening browser tabs (open_url).
- Sanitization: Relies on user verification of the scoped procedure and stages before script creation.
Audit Metadata