clean-docs

SKILL.md

Use planning-with-files skill with reset. The task:

Task

Scan every .mdx doc in www/content/docs/ and clean it following @docs/solutions/style.md.

Rules

KEEP (do not remove)

  • Callouts (<Callout>) — especially at end of sections
  • Intro/transitioning text at start of docs and sections ("In this guide...", "Let's...", "We'll...")
  • Conversational tone, progressive build-up, celebratory endings
  • All technical content — be lossless on information

REMOVE

  • Marketing filler: "battle-tested", "secret sauce", "you can copy into your project"
  • True redundancy: same info stated twice in same doc (duplicate code blocks, sentences restating a table above)
  • DRY violations: identical code blocks appearing 2-3x in same file — keep first occurrence, replace later ones with cross-reference ("See section above")
  • Technical details users don't care about: internal implementation notes that don't help usage

FIX

  • Linear ordering: concepts should build progressively, not jump back and forth

Process

  1. Use planning-with-files to create task_plan.md with all 54 docs as checklist items
  2. Batch docs into parallel agents (5-6 files per agent) for scanning
  3. Pass 1 — conservative: grep for known filler patterns, fix obvious marketing language
  4. Pass 2 — deeper: read each file, find duplicate code blocks, DRY violations, ordering issues
  5. Mark each doc as — no change needed or — updated: [what changed]
  6. Run bun lint:fix after all edits
Weekly Installs
1
GitHub Stars
378
First Seen
3 days ago
Installed on
amp1
cline1
opencode1
cursor1
kimi-cli1
codex1