Claude Code Plugins

Community-maintained marketplace

Feedback

lint-markdown

@CsHeng/dot-claude
5
0

Execute markdown validation with taxonomy-based classification and custom rules. Use when validating markdown compliance with LLM-facing writing standards or when generating structured validation reports.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name lint-markdown
description Execute markdown validation with taxonomy-based classification and custom rules. Use when validating markdown compliance with LLM-facing writing standards or when generating structured validation reports.
allowed-tools Bash(python3), Read, Glob, Grep

Purpose

Execute Python-based markdown validation with three-tier classification based on taxonomy-rfc.md: STRICT files require full compliance with LLM-facing standards, MODERATE files apply governance rules, and LIGHT files receive basic validation.

IO Semantics

Input: File paths, directories, or global workspace scope with optional parameters.

Output: Structured linting reports with issue categorization, severity levels, and auto-fix suggestions when applicable.

Side Effects: Updates target files when using --fix parameter, generates structured reports in JSON or human-readable format.

Deterministic Steps

1. Environment Validation

  • Verify Python 3 availability.
  • Confirm validator script exists at skills/llm-governance/scripts/validator.py.
  • Validate config.yaml exists and loads properly.

2. File Classification

  • Apply STRICT classification to LLM-facing files: commands//*.md, skills//SKILL.md, agents//AGENT.md, rules//*.md, AGENTS.md, CLAUDE.md
  • Apply MODERATE classification to governance files: governance//*.md, config-sync//.md, agent-ops/**/.md
  • Apply LIGHT classification to remaining markdown files.
  • Exclude human-facing docs: docs/, examples/, tests/, ide/

3. Validation Execution

  • Run Python validator based on requested mode: python3 skills/llm-governance/scripts/validator.py for standard validation python3 skills/llm-governance/scripts/validator.py for JSON output (future)
  • Parse validator output and categorize issues by severity and type.

4. Report Generation

  • Aggregate results by file classification and issue type.
  • Generate structured summary with:
    • Total issue count and severity breakdown
    • Classification-specific compliance metrics
    • Auto-fix success rate where applicable
  • Provide actionable recommendations organized by priority.

5. Validation Compliance

  • Ensure all processing respects skills/llm-governance/rules/99-llm-prompt-writing-rules.md constraints.
  • Apply imperative communication patterns in all output.
  • Maintain 100-character line limits in generated reports.

Safety Constraints

  • Never modify files without explicit --fix parameter.
  • Preserve original file content through backup mechanisms when fixing.
  • Respect file exclusions and never scan excluded directories.
  • Validate tool chain compatibility before executing validator.