Connect Saves, CLI Commands, Tasks, and CI

Connect language checks through Bash, VS Code Tasks, and CI, and diagnose version, path, and configuration mismatches.

Sources checked on 2026-09-06. Apply the examples sequentially in a separate format-lab project. Commands run from its root on macOS, Linux, or WSL. Merge the relevant keys into existing configuration files.

Finish with one checking entry point

Saving handles the current file; CLI checks cover the agreed repository scope. Share commands that read the same configuration instead of attempting to replay editor hooks in CI.

Entry point	Responsibility	Success
Editor save	Language formatter and selected fixes	Saved file passes CLI checks
CLI check	Agreed repository scope	Every command exits zero
VS Code Task	Invoke CLI check	Same terminal output and status
CI	Install pinned environment and run check	Pass from a clean checkout

Create a checking script

This requires the files from all prerequisite lab pages. Save as scripts/check.sh. It changes to the lab root relative to its own location, making configuration lookup independent of the caller's directory.

#!/usr/bin/env bash
set -euo pipefail
cd "$(dirname "${BASH_SOURCE[0]}")/.."

npm run fmt:check
npm run lint
npm run typecheck
uv run --locked yamllint -c .yamllint.yaml -s config
shfmt -ln bash -i 2 -d scripts
shellcheck scripts/*.sh
for script in scripts/*.sh; do
  bash -n "$script"
done
bash scripts/check-go-format.sh
(cd go-example && go vet ./... && go test ./...)
uv run --locked ruff format --check python-example
uv run --locked ruff check python-example

The script stops at the first failure. Go runs inside a subshell so Python commands still use the root. Include check-go-format.sh from Go formatting.

Keep write and fix options out of this entry point. The lab scopes YAML and Python checks to example directories, ShellCheck to immediate scripts, and Oxc to its configured exclusions. Update those scopes when adding directories.

This script does not validate the YAML application schema or perform complete Python type checking. Add those commands when required. Running Go tests with no test files is also not a behavioral test suite.

Apply fixes separately

First complete each example's manual corrections. Run these commands individually; resolve remaining diagnostics before continuing.

npm run lint:fix
npm run fmt
shfmt -ln bash -i 2 -w scripts
gofmt -w go-example
uv run --locked ruff check --fix python-example
uv run --locked ruff format python-example
bash scripts/check.sh
git diff --check
git diff

The first pass can also format configuration files and scripts. Review the diff, repeat the commands, and confirm no additional changes. git diff --check checks whitespace errors and does not replace language tools.

Let a VS Code Task call the entry point

Append this object to the tasks array from YAML:

{
    "label": "check",
    "type": "process",
    "command": "bash",
    "args": ["scripts/check.sh"],
    "options": { "cwd": "${workspaceFolder}" },
    "group": { "kind": "test", "isDefault": true },
    "problemMatcher": []
}

Run Tasks: Run Task → check. This task uses terminal output; use YAML's dedicated YAML task for navigable YAML diagnostics. A process task and argument array avoid constructing shell command strings. VS Code Tasks

Run the same commands in CI

Provision Node.js/npm, uv/Python, the Go SDK, shfmt, and ShellCheck first. These are the commands after tool installation, not a complete runner provisioning workflow:

npm ci
uv sync --locked
bash scripts/check.sh

Commit both lockfiles. Pin Node, uv, Go, shfmt, and ShellCheck through the runner image or tool version configuration too. Resolving latest tools on every run cannot be made reproducible by JavaScript and Python lockfiles alone.

Compare these outputs locally and in CI:

node --version
npm --version
npm exec -- oxfmt --version
npm exec -- oxlint --version
uv --version
uv run --locked ruff --version
uv run --locked yamllint --version
go version
shfmt --version
shellcheck --version

Diagnose differences in order

Symptom	Inspect	Action
Saving and CLI repeatedly undo changes	Competing format and import actions	Assign language-specific providers
Only the terminal finds a tool	Extension host PATH	Inspect Output and executable path
Only a subfolder differs	cwd, nested configs, ignores	Compare discovery and file scope
Local passes, CI fails	Versions, locks, untracked files	Reproduce from a clean checkout
Only a remote window fails	Local versus remote execution	Install tools and extensions remotely

Keep personal absolute paths and OS-specific settings separate when sharing workspace configuration. Extension recommendations guide installation; they do not install CLI tools or enforce policies.

Final exercise

Reintroduce TS any, a YAML duplicate key, unquoted Shell expansion, a mismatched Go Printf verb, and an unused Python import one at a time. Record which command fails and which still passes. After fixing them, compare save → check Task → CLI check → clean-checkout CI check.

Sources were checked against official documentation and configuration contracts. Verify actual editor save actions, Problems navigation, and remote extension behavior in your environment using the exercises. A passing CLI check is not evidence that those GUI interactions were tested.

CLI verification record

On 2026-09-06, the code blocks were extracted into a temporary project and exercised with Oxfmt 0.66.0, Oxlint 1.81.0, shfmt 3.12.0, ShellCheck 0.9.0, Go 1.27.1, Ruff 0.16.6, and yamllint 1.38.0. Oxc and TypeScript reused the workspace installation; Python tools were installed separately through uv.

Checks covered expected failures, corrected examples, Shell output with spaces, the yamllint diagnostic regex, the combined check.sh, and no additional changes on a second format pass. VS Code GUI interactions and an actual CI runner were outside this verification.