{"id":"maxritter-pilot-shell","name":"pilot-shell","af_score":43.2,"security_score":47.8,"reliability_score":36.2,"what_it_does":"Pilot Shell is an installation/CLI wrapper around Claude Code that provides a structured “spec-driven development” workflow (plan/approve/implement/verify), enforced quality hooks (lint/format/typecheck), git worktree-based task isolation, and a local dashboard for session/spec/review/status visibility.","best_when":"You are already using Claude Code and want stronger engineering discipline (TDD, verification, review gates) with a consistent workflow across sessions and projects.","avoid_when":"You need a simple chat interface only, or you cannot accommodate local installation, hooks, and test execution as part of the agent loop.","last_evaluated":"2026-03-30T13:23:26.312751+00:00","has_mcp":true,"has_api":false,"auth_methods":["Claude subscription/auth via claude auth status (as referenced in README for tier detection)"],"has_free_tier":false,"known_gotchas":["Requires local environment setup (dependencies, language servers, hooks); failures may occur if prerequisites can’t be installed/initialized.","Worktree-based implementation and browser automation for UI E2E may be slow/fragile in constrained environments.","Headless mode changes interaction style; agent assumptions about interactive approval may not hold without proper flags."],"error_quality":0.0}