Add per-sample skill resolution + reproducibility stamping #128

	name: CI_tests

	on:
	pull_request:
	branches: ['main']

	jobs:
	ci_test:
	runs-on: ubuntu-latest
	env:
	IS_CI: true
	OPENAI_API_KEY: "Fake key, but the var needs to be set or else Inspect will complain at import time"
	HF_TOKEN: ${{ secrets.HF_TOKEN }}
	steps:
	- uses: actions/checkout@v4

	# Set up Docker Buildx for caching
	- name: Set up Docker Buildx
	uses: docker/setup-buildx-action@v3

	# Build the Docker image with layer caching
	- name: Build main Docker image
	uses: docker/build-push-action@v5
	with:
	context: .
	file: ./docker/Dockerfile
	tags: agent-baselines:latest
	target: agent-baselines
	cache-from: type=gha,scope=agent-baselines # Use GitHub Actions cache
	cache-to: type=gha,mode=max,scope=agent-baselines
	load: true # Make the image available for later steps

	- name: Run tests
	run: \|
	make test

	- name: Smoke solvers (per-solver envs)
	run: \|
	# Keep this small but meaningful: exercise two solvers with different inspect_ai versions.
	make smoke-solvers-ci

Provide feedback