The first packaged release of The Wizard's Pick.
- Provides manual, assisted, and automated execution modes for authorized security testing.
- Records session scope, command history, findings, and Markdown reports.
- Uses project-local Ollama inference by default with a pinned DeepHat-V1-7B model artifact.
- Adds bounded command capture, process-group cleanup, private local state, and literal report encoding.
- Adds Python 3.11 through 3.14 CI, CodeQL, Dependabot, and trusted PyPI publishing.
Automated mode intentionally executes parsed model proposals without confirmation. Pick does not include a scope matcher, command denylist, sandbox, approval service, or output redaction. Operators provide those controls through their execution environment and engagement process.