heymegabyte/agent-skills / 09-brand-and-content-system
brand-and-content-system
VerifiedExtract real brands (Wayback for rebuilds). Copy system, headline/CTA rules, trust surfaces, legal pages, SEO+structured data, anti-AI-slop, microcopy, DESIGN.md, W3C DTCG tokens, pSEO 5 types, GEO/AI search.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 12.2 KB |
| ai-search-geo.md | markdown | 4.0 KB |
| build-breaking-rules.md | markdown | 21.0 KB |
| documentation-and-codebase-hygiene.md | markdown | 3.1 KB |
| email-templates.md | markdown | 6.0 KB |
| grammar-audit.md | markdown | 3.7 KB |
| per-route-metadata.md | markdown | 6.8 KB |
| seo-and-keywords.md | markdown | 6.3 KB |
| social-automation.md | markdown | 5.8 KB |
grammar-audit.md
View on GitHub---
name: "grammar-audit"
description: "Final LLM-powered grammar/clarity pass after copy generation. Fixes typos, agreement, tense, capitalization, Oxford comma WITHOUT rewriting or removing information. Outputs a diff, not a rewrite."
updated: "2026-05-01"
---
# Grammar Audit (***FINAL COPY PASS — EVERY SITE, EVERY BUILD***)
After all copy is generated (hero, body, blog, FAQ, microcopy, legal), run `grammar-audit.mjs` (GPT Image 2 vision-mini, concurrency 5-8) to fix mechanical errors WITHOUT rewriting — a corrective filter, not a creative rewrite.
## Mandate (NON-NEGOTIABLE)
### Fix
- Typos
- Spelling
- Subject-verb agreement
- Tense consistency
- Pronoun agreement
- Apostrophes / possessives
- Hyphens vs en / em dashes
- Capitalization (proper nouns, titles, sentence case)
- Oxford comma consistency (apply Oxford by default — flip site-wide if voice rejects it)
- Quotation mark direction
- Spacing (single space after period)
- Numbers (≥10 numeric, <10 spelled — except dates / times / money)
- a vs an before vowel sounds
- than vs then
- who vs whom (when natural)
- its vs it's
- their / there / they're
- affect vs effect
### NEVER
- Rewrite for style
- Shorten
- Lengthen
- Change meaning
- Remove content
- Replace banned-word with synonym (anti-slop is skill 09's job, not grammar's)
- Rephrase voice
- Add / remove information
- Reformat structure
- Translate idioms
## Output Shape
```json
{
"edits": [
{ "file": "src/data/blog-posts.ts", "before": "She recieved her PHD in 2019,", "after": "She received her PhD in 2019,", "reason": "spelling+capitalization" },
{ "file": "src/pages/about.tsx", "before": "...colleagues, mentors and students.", "after": "...colleagues, mentors, and students.", "reason": "oxford-comma" }
]
}
```
Apply edits in-place via `Edit` tool calls. Idempotent — re-running on already-corrected copy produces empty `edits[]`.
## Run Conditions
1. After `clean_content` blog import (skill 15) and after `enhance-blog-posts.mjs`
2. After main page copy generation (homepage, about, services, contact, FAQ)
3. After legal page copy (privacy, terms, accessibility, cookies)
4. Before `gorgeous-loop` final pass (skill 06)
5. Re-run on demand via `node scripts/grammar-audit.mjs --only=<file>`
## Concurrency + Cost
- GPT Image 2 vision-mini at ~$0.15/1M input + $0.60/1M output
- Per site — ~50K input tokens (all copy concatenated) + ~5K output (diff only) = ~$0.01/site
- Concurrency 5-8 against the API
- CLI — `--only=path`, `--limit=N`, `--dry-run`
## Scope Boundaries (***WHAT IT NEVER TOUCHES***)
- Direct quotes (preserve verbatim — author voice)
- Named-entity spelling (verified against `_research.json`)
- Brand names
- Technical jargon (verified against `_research.json.industry_terms`)
- Code blocks
- URLs
- Email addresses
- Phone numbers
- Dates
- Money figures
- Proper nouns flagged in `_research.json.entities[]`
## Hard Gate
After grammar pass re-run `gate` script. If it introduced any error caught by other gates (link 404, citation regex match, banned-word match), revert that specific edit and flag in `audit_logs` — grammar pass never breaks other gates.
## Voice Preservation
Author voice is sacred. Never normalize "ain't" → "is not" for colloquial voice, remove rhetorical fragments, or add "that" when omitted intentionally ("She knew [that] he'd come."). The audit asks — "Did the author intend this?" → if yes, leave alone.
## Reference Anchors
- Strunk & White (1959) for clarity baseline
- Chicago Manual of Style 17th ed for capitalization / comma
- APA 7th for academic citations
- AP Stylebook for journalism
**Default:** Chicago for body copy, AP for blog / news, APA for citations (`rules/citations.md`).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |