sf-eval-starter

You can automate the RED GREEN REFACTOR TDD loop by orchestrating SpecFormula MCP goals, delegating phase execution to mapped sub-skills, and polling verifier runs every 20 seconds until completion.

SpecFormula-driven TDD evaluation fetches the next TDD goal, builds a deterministic non-globbed test command covering current and completed feature files, dispatches it via MCP, and polls verify_goal_status until the all-done sentinel is returned.

Yes, automated TDD orchestration works with any compatible test runner capable of executing explicit .feature scopes, requiring correct project routing via project_path and mapped aibdd-red-execute, aibdd-green-execute, and aibdd-refactor-execute sub-skills.

Automated test evaluation fails cleanly when the local preflight dry-run cannot launch the test runner in the worker context, preventing invalid or environment-incompatible test commands from being dispatched.

Yes, MCP tool availability is strictly required to fetch goals, dispatch SpecFormula jobs, and poll verify_goal_status every 20 seconds until the evaluator returns the completed verdict.

To ensure deterministic test evidence, construct an explicit, non-globbed test command that includes the current goal feature file plus all previously completed DONE feature files before launching the runner.