Markdown linter for agent output quality

Summary

Integrate a Markdown linter so DocsClaw agents can validate and clean up their own Markdown output.

Motivation

DocsClaw agents often produce Markdown documents (summaries, reports, converted content). A built-in linting step ensures consistent, well-formed output without relying on the LLM to get formatting right every time.

Implementation options

1. CLI tool in the image — bundle markdownlint-cli and invoke via exec as a post-processing step
2. Built-in tool — add a lint_markdown tool that runs markdownlint on a file and returns violations
3. Skill-based — provide a SKILL.md that instructs the agent to lint its own output before finalizing

Option 2 is probably the most useful — the agent can lint, see violations, and fix them in the agentic loop.

Acceptance criteria

• Agent can lint Markdown files it produces
• Linting rules configurable via .markdownlint.yaml in the config directory
• Works in the docs and content image flavors at minimum

Start by comparing the three implementation options, with particular attention to the proposed lint_markdown tool and the existing config directory. Check how the docs and content image flavors are defined and where markdownlint-cli could run. Done means Markdown files can be linted, .markdownlint.yaml rules are honored, and both flavors support the workflow.