wizard

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is instructed to read various repository files to identify configuration needs, which could be a surface for indirect instructions.
  • Ingestion points: Processes project files including .env, .env.example, README.md, docker-compose.yml, and .github/workflows/* to determine configuration requirements.
  • Boundary markers: No specific delimiters or instruction-blocking markers are defined for these inputs.
  • Capability inventory: The skill generates executable bash scripts capable of file writing (write_env) and network operations via the GitHub CLI (set_secret).
  • Sanitization: The instructions explicitly require the agent to present the planned stages and values to the user for confirmation and modification before the script is authored.
  • [COMMAND_EXECUTION]: The skill's primary purpose is the generation and subsequent execution of bash scripts (chmod +x <script>) on the user's local machine.
  • [DATA_EXFILTRATION]: The generated scripts are designed to capture user-provided credentials and exfiltrate them to the user's own GitHub repository secrets using the gh command-line tool. This is a legitimate and documented function of the skill for CI/CD setup.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 11:39 PM
Security Audit — agent-trust-hub — wizard