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