AgentSkillsCN

md-normalizer

利用 skills/md-normalizer/scripts/md_normalizer.rb，结合可选的 Pandoc 备用方案，将 Markdown 或 HTML 快照归一化为规范的 Markdown 格式。当您在章节拆分或内容提取前需要确保归一化过程具有确定性时，此方法尤为实用。

SKILL.md
--- frontmatter
name: md-normalizer
description: Normalize Markdown or HTML snapshots into canonical Markdown using skills/md-normalizer/scripts/md_normalizer.rb with optional pandoc fallback. Use when you need deterministic normalization before section splitting or extraction.

MD Normalizer

Overview

Normalize snapshot content (Markdown or HTML) into stable GitHub-flavored Markdown and write metadata/state updates.

Quick Start

bash
# list sources and their last snapshot path
skills/md-normalizer/scripts/md_normalizer.rb list

# normalize all sources in state.json
skills/md-normalizer/scripts/md_normalizer.rb normalize --all

# normalize a single source by id
skills/md-normalizer/scripts/md_normalizer.rb normalize --id best-practices

Inputs

  • •data/doc-fetcher/state.json (last snapshot path per source)
  • •Snapshot files under data/doc-fetcher/snapshots/<id>/
  • •Subcommands: list, normalize
  • •Optional flags for normalize: --all, --id, --force, --dry-run

Outputs

  • •Normalized Markdown: data/doc-fetcher/normalized/<id>/<sha>.md
  • •Normalization metadata: data/doc-fetcher/normalized/<id>/<sha>.json
  • •State updates: data/doc-fetcher/state.json

Workflow

  1. •Ensure snapshots exist (run doc-fetcher first).
  2. •Run normalization with normalize --all or normalize --id.
  3. •Confirm normalized output and state updates.

Options

  • •list: Print sources and last snapshot path (from state.json).
  • •normalize --force: Overwrite existing normalized output.
  • •normalize --dry-run: Do not write files.

Notes

  • •Prefers .md snapshot passthrough; uses pandoc for HTML.
  • •A wrapper exists at scripts/md_normalizer.rb.