fix: Makefile interpreter consistency + setuptools package discovery

- Single PYTHON variable (default /usr/local/opt/python@3.11/bin/python3.11)
used for all targets via $(PYTHON) -m pip/pytest/uvicorn
- Explicit package discovery: [tool.setuptools.packages.find] include=["agent_bench*"]
fixes editable install when non-package dirs (docs/plans) exist
- Relax types-PyYAML version constraint for cross-environment compat

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Makefile

@@ -1,10 +1,12 @@
+PYTHON ?= /usr/local/opt/python@3.11/bin/python3.11
+
 .PHONY: install test lint serve ingest evaluate-fast evaluate-full benchmark docker
 
 install:
-	pip install -e ".[dev]"
+	$(PYTHON) -m pip install -e ".[dev]"
 
 test:
-	pytest tests/ -v --tb=short
+	$(PYTHON) -m pytest tests/ -v --tb=short
 
 lint:
 	ruff check agent_bench/ tests/
@@ -12,19 +14,19 @@ lint:
 	mypy agent_bench/ --ignore-missing-imports
 
 serve:
-	uvicorn agent_bench.serving.app:create_app --factory --reload --port 8000
+	$(PYTHON) -m uvicorn agent_bench.serving.app:create_app --factory --reload --port 8000
 
 ingest:
-	python3 scripts/ingest.py --config configs/tasks/tech_docs.yaml
+	$(PYTHON) scripts/ingest.py --config configs/tasks/tech_docs.yaml
 
 evaluate-fast:
-	python3 scripts/evaluate.py --config configs/default.yaml --mode deterministic
+	$(PYTHON) scripts/evaluate.py --config configs/default.yaml --mode deterministic
 
 evaluate-full:
-	python3 scripts/evaluate.py --config configs/default.yaml --mode full
+	$(PYTHON) scripts/evaluate.py --config configs/default.yaml --mode full
 
 benchmark:
-	python3 scripts/benchmark.py --output docs/benchmark_report.md
+	$(PYTHON) scripts/benchmark.py --output docs/benchmark_report.md
 
 docker:
 	docker-compose -f docker/docker-compose.yaml up --build

pyproject.toml

@@ -27,9 +27,12 @@ dev = [
     "ruff>=0.6.0",
     "mypy>=1.11.0",
     "respx>=0.21.0",
-    "types-PyYAML>=2024.0.0",
+    "types-PyYAML",
 ]
 
+[tool.setuptools.packages.find]
+include = ["agent_bench*"]
+
 [build-system]
 requires = ["setuptools>=69.0"]
 build-backend = "setuptools.build_meta"