ci: add markdown-lint and link-check CI (INFRA-128)

Add a standalone ci workflow so handbook changes get an automated gate.
The job is named literally "ci / ci" to emit the exact status context the
org main-branch-protection ruleset requires.

- markdownlint-cli2 (.markdownlint-cli2.jsonc): MD013/MD060/MD040 relaxed
  as noisy docs-style rules; fixed 3 MD032 blank-line-around-list issues.
- lychee link check (lychee.toml): internal + external links, tolerates 429.

.github/workflows/ci.yaml

@@ -0,0 +1,52 @@
+name: ci
+
+# CI for the engineering handbook. This repo is docs-only (Markdown), so the
+# gate lints Markdown and checks that links resolve.
+#
+# Naming is load-bearing: the org "main branch protection" ruleset matches the
+# required status check against the JOB's check-run name, NOT "workflow / job".
+# For a normal (non-reusable) job the check-run name IS the job name, so the job
+# must be named literally "ci / ci" to emit that exact context. (A job named
+# "ci" emits the context "ci" — which the PR UI cosmetically displays as
+# "ci / ci" but does NOT satisfy the requirement.) This mirrors the org's
+# aggregator-job convention in the `.github` repo's own self-CI.
+#
+# Both tools are configured to be green on the current content:
+#   - markdownlint-cli2 reads .markdownlint-cli2.jsonc
+#   - lychee reads lychee.toml (tolerates 429 so external flakiness never reds
+#     an otherwise-valid docs change; broken internal links still fail)
+
+on:
+  pull_request:
+  push:
+    branches: [main]
+
+permissions:
+  contents: read
+
+jobs:
+  ci:
+    name: ci / ci
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    concurrency:
+      group: ci-${{ github.workflow }}-${{ github.ref }}
+      cancel-in-progress: true
+    steps:
+      - uses: actions/checkout@v6
+
+      - uses: actions/setup-node@v6
+        with:
+          node-version: "24"
+
+      - name: Markdown lint
+        run: npx --yes markdownlint-cli2@0.23.0
+
+      - name: Link check
+        uses: lycheeverse/lychee-action@8646ba30535128ac92d33dfc9133794bfdd9b411 # v2.8.0
+        with:
+          args: "--config lychee.toml --no-progress './**/*.md'"
+          fail: true
+        env:
+          # Authenticated github.com requests avoid API rate-limit 429s.
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.markdownlint-cli2.jsonc

@@ -0,0 +1,23 @@
+{
+  // markdownlint-cli2 config for the engineering handbook.
+  // Run locally with: npx --yes markdownlint-cli2 "**/*.md" "#node_modules"
+  // CI runs the same command (see .github/workflows/ci.yaml).
+  "config": {
+    "default": true,
+
+    // MD013 (line-length): disabled. The handbook is prose and wide reference
+    // tables; hard-wrapping at 80 columns hurts readability and diffs. Line
+    // length is not a correctness concern for docs.
+    "MD013": false,
+
+    // MD060 (table-column-style): disabled. Purely cosmetic pipe-padding style
+    // for compact tables; the existing tables are compact and render fine.
+    "MD060": false,
+
+    // MD040 (fenced-code-language): disabled. Several fenced blocks are
+    // directory trees / plain-text output that have no meaningful language tag.
+    "MD040": false
+  },
+  "globs": ["**/*.md"],
+  "ignores": ["node_modules", ".git"]
+}

aws-infrastructure.md

@@ -31,6 +31,7 @@ aws-cdk-lib bundles transitive dependencies (`inBundle: true`) that npm `overrid
 If a specific release is known-bad, ignore that version only (`ignore: - dependency-name: aws-cdk-lib, versions: ["2.254.0"]`) with a comment explaining why, and remove the entry once a fixed release ships.
 
 When upgrading, verify on a branch first:
+
 1. Update `package.json` to the new version
 2. Run `rm -rf node_modules package-lock.json && npm install`
 3. Run `npm ci` — if it fails, the version is not safe

github-standards.md

@@ -26,6 +26,7 @@ Every active repo runs the same baseline. The security half is meant to be carri
 | CodeQL default setup | configured | per-repo (`PATCH .../code-scanning/default-setup state=configured`) until added to the config |
 
 Exceptions:
+
 - **Docs-only repos** (e.g. `engineering-handbook`) skip CodeQL — there is no compiled code to scan; `code_security` may stay off. Secret scanning still applies.
 - **`shoc-backend` / `shoc-frontend-new`** are excluded from org compliance tooling (see the `.github` org-config notes); leave their settings to the SHOC team.
 

lychee.toml

@@ -0,0 +1,20 @@
+# lychee link-checker config for the engineering handbook.
+# Used by CI (lychee-action in .github/workflows/ci.yaml) and reproducible
+# locally with:
+#   docker run --rm -v "$PWD:/input" lycheeverse/lychee \
+#     --config /input/lychee.toml /input/**/*.md
+
+# Cache results between runs to cut repeat network calls.
+cache = true
+
+# Retry transient failures before giving up.
+max_retries = 2
+retry_wait_time = 2
+timeout = 20
+
+# Treat 2xx as success and tolerate 429 (rate-limited) so external flakiness
+# never reds an otherwise-valid docs change. Broken internal links still fail.
+accept = ["200..=299", "429"]
+
+# Don't traverse dependency dirs.
+exclude_path = ["node_modules", ".git"]

secrets-and-config.md

@@ -23,6 +23,7 @@ stack-name/value-name
 ```
 
 Examples:
+
 - `my-stack/slack-signing`
 - `my-stack/stripe-key`