diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
index 098f18f0e..dfca2c851 100644
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@@ -10,7 +10,7 @@ on:
   push:
     branches:
       - main
-    
+
 name: build
 
 jobs:
@@ -24,7 +24,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-node@v2
         with:
-          node-version: '17'  
+          node-version: "17"
       - run: cd ./tntc && npm install
       - run: cd ./tntc && npm run compile
       - run: cd ./tntc && npm test
@@ -39,11 +39,17 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-node@v2
         with:
-          node-version: '17'  
+          node-version: "17"
       - run: npm install -g txm
       - run: cd ./tntc && npm install
       - run: cd ./tntc && npm run compile && npm link
-      - run: cd ./tntc && txm cli-tests.md
+      - name: Blackbox integration tests
+        run: cd ./tntc && txm cli-tests.md
+      - name: Blackbox integration tests with I/O
+        # This tests fail on windows currently
+        # See https://github.com/anko/txm/issues/10
+        run: cd ./tntc && txm io-cli-tests.md
+        if: matrix.operating-system != 'windows-latest'
 
   tntc-antlr-grammar:
     runs-on: ubuntu-latest
@@ -51,7 +57,6 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-node@v2
         with:
-          node-version: '17'  
+          node-version: "17"
       - run: cd ./tntc && npm install
       - run: cd ./tntc && npm run antlr
-
diff --git a/tntc/runner.sh b/tntc/runner.sh
deleted file mode 100755
index ecf6890d6..000000000
--- a/tntc/runner.sh
+++ /dev/null
@@ -1,33 +0,0 @@
-#!/usr/bin/env bash
-
-# This runner script just reads each line from stdin and executes it, stripping off any
-# trailing `\r` to normalize output on windows.
-#
-# The runner script ensures that the "program" run by txm is consistent between OSs:
-# I.e. that it is always run with bash
-#
-# We use it in ./cli-tests.md to run our black box CLI commands.
-
-set -euo pipefail
-
-export LANG="C.UTF-8"
-export LANGUAGE=
-export LC_CTYPE="C.UTF-8"
-export LC_NUMERIC="C.UTF-8"
-export LC_TIME="C.UTF-8"
-export LC_COLLATE="C.UTF-8"
-export LC_MONETARY="C.UTF-8"
-export LC_MESSAGES="C.UTF-8"
-export LC_PAPER="C.UTF-8"
-export LC_NAME="C.UTF-8"
-export LC_ADDRESS="C.UTF-8"
-export LC_TELEPHONE="C.UTF-8"
-export LC_MEASUREMENT="C.UTF-8"
-export LC_IDENTIFICATION="C.UTF-8"
-export LC_ALL=
-
-while IFS= read -r cmd; do
-    # All the noise here with FDs is to filter both stdout and stderr
-    # See https://stackoverflow.com/a/31151808/1187277
-    bash -c "$cmd"
-done