Makefile

.DEFAULT_GOAL := help
.PHONY: help demo gym agent agent-replay dev-setup install typecheck lint format depcruise test coverage drift verify clean \
        mutate-baseline mutate-baseline-pathological mutate-replay mutate-all report semantic

help: ## Show this menu
	@awk 'BEGIN {FS = ":.*?## "} \
		/^##@/ {printf "\n\033[1m%s\033[0m\n", substr($$0, 5)} \
		/^[a-zA-Z_-]+:.*?## / {printf "  \033[36m%-22s\033[0m %s\n", $$1, $$2}' $(MAKEFILE_LIST)

##@ Start here

dev-setup: install ## Install dependencies and prepare the workspace
	@echo "Workspace ready. Run 'make demo' for the whole lab, end to end."

demo: ## The whole lab: instrument arms, gap gym, flagship audit, report. No API keys
	pnpm demo

##@ The lab, one act at a time

gym: ## Is the agent's learning real? Three policies vs planted blind spots, no API keys
	pnpm gym

agent-replay: ## Audit the committed flagship agent run byte for byte, no API keys
	pnpm agent:replay runs/flagship-seed42

agent: ## Run the agent live (the ONLY paid target; needs .env keys, stops at TMUTATE_LLM_BUDGET_USD)
	pnpm agent

report: ## Rebuild the scored writeup and site/index.html from run outputs
	pnpm report

##@ Development gates

verify: ## Typecheck, lint, layering gate, and tests
	pnpm verify

test: ## Run the test suite once
	pnpm test

coverage: ## Run the test suite with coverage floors
	pnpm test:coverage

drift: ## Replay the committed recordings against the current build (does the code still match the tape?)
	pnpm drift

typecheck: ## Run the TypeScript type checker
	pnpm typecheck

lint: ## Run ESLint
	pnpm lint

format: ## Format the codebase with Prettier
	pnpm format

depcruise: ## Run the dependency layering gate
	pnpm depcruise

##@ Instrument arms and extras

mutate-baseline: ## The conventional mutation tester, for comparison
	pnpm mutate:baseline

mutate-replay: ## The replay method, workflow tier only
	pnpm mutate:replay

mutate-all: ## The replay method across all tiers
	pnpm mutate:replay:all

mutate-baseline-pathological: ## The baseline against a real time server, the worst case setup
	pnpm mutate:baseline:pathological

semantic: ## Judge calibration then the standalone adversary (needs .env keys)
	pnpm calibrate && pnpm adversary

install: ## Install dependencies with pnpm
	pnpm install

clean: ## Remove build and cache artifacts
	rm -rf dist coverage .stryker-tmp .vitest reports