[Bug]: markdownlint CI job lints 0 files — glob quoting makes lint.yml a no-op

Bug Description

The markdownlint job in .github/workflows/lint.yml lints zero files on every run, so it reports success unconditionally. Markdown formatting has not been enforced in CI since 2026-02-10.

Steps to Reproduce

1. Open Lint workflow run 34513299255 for commit c173bf19a6654e3b05386ec3599349a55282b897 and inspect the markdownlint job. It reports Linting: 0 files and Summary: 0 issues in 0 files.
2. Reproduce locally with: npx --yes [email protected] "'/*.md'" "!extensions//*.md".
3. Remove the single quotes; the command then matches files and reports the existing violations.

Expected Behavior

The job should lint the tracked Markdown files it is intended to cover (/*.md except extensions//*.md) and fail when they contain lint errors.

Actual Behavior

The job matches no files and always passes.

Specify CLI Version

Not applicable — CI workflow defect at main @ c173bf19a6654e3b05386ec3599349a55282b897

AI Agent

DeepSeek Harness

Operating System

Windows 11 host; behavior confirmed in the live ubuntu-latest runner log

Python Version

Not applicable — workflow defect

Error Logs

markdownlint-cli2 v0.23.2 (markdownlint v0.41.1)
Finding: '**/*.md' !extensions/**/*.md
Linting: 0 files
Summary: 0 issues in 0 files

Additional Context

Root cause: the YAML block scalar in .github/workflows/lint.yml preserves the single quotes around **/*.md, so markdownlint-cli2 receives a quoted glob that matches nothing. The issue was introduced by commit f14a47e. Running the intended scope against tracked files reports 277 issues across 32 files. Removing the quotes would make CI meaningful but red, so maintainer direction is needed for the intended scope and cleanup plan.

AI Assistance Disclosure: AI assistance was used for repository inspection, reproduction, evidence gathering, and drafting this report. The finding was checked against the frozen commit, live CI log, local reproduction output, and all-state issue/PR searches; no code changes were made for this report. Related issue #1030 does not report this exact defect.

Specific model and mode: OpenAI Codex (GPT-5), autonomous mode.