draft-article

draft-article

Generate a complete long-form article (1500–2500 words) based on the article plan. The article includes inline code snippets, direct links to repository files, and is grounded in verified claims. Output is publish-ready markdown.

When to use

Run after /linkedin-content:article-plan. This produces the actual article — the previous skill only produced the outline.

User input

$ARGUMENTS

Treat user input (if any) as additional tone, audience, or focus overrides.

Prerequisites

Verify these files exist:

1. content/article-plan.md
2. content/claims.md

If article-plan.md is missing, output and STOP:

ERROR: Missing prerequisite.
- Article plan not found at content/article-plan.md
- Run /linkedin-content:article-plan first to generate it.

If claims.md is missing, output and STOP:

ERROR: Missing prerequisite.
- Claims not found at content/claims.md
- Run /linkedin-content:proof-pack first to generate it.

Process

Step 1: Load context

1. Read content/article-plan.md — extract title, subtitle, metadata, and every section's plan (key points, target word count, claims, repo links, code snippet indicators).
2. Read content/claims.md — load all claims for inline referencing.
3. Read every source file referenced by "Code snippet" entries — extract the verbatim excerpts to embed.

Step 2: Optional config override

If content/.config.md exists, read it. It may override defaults like:

• Forbidden phrases — additional banned phrases beyond the defaults below
• Hashtag set — required/preferred hashtags
• Tone — voice instructions (e.g., "warm and conversational", "terse and technical")

If absent, use defaults.

Step 3: Write the article

Follow the plan's section order and headings. The article MUST:

• Be 1500–2500 words total (count carefully)
• Hit each section's target word count (±20%)
• Include every key point listed in the plan for each section
• Embed inline code snippets where the plan indicates — fenced code blocks with language tags (e.g., ```python, ```typescript, ```bash, ```apex)
• Include direct repo links using full URLs from the plan
• Reference verified claims — do not invent capabilities or stats not in claims.md
• Use no forbidden phrases: "AI is changing everything", "the future of", "game-changer", "revolutionary", "next-generation" — plus any additional phrases from .config.md
• Maintain a technical-but-accessible tone — precise enough for developers, clear enough for technical managers
• Include a TL;DR block at the top (3–4 bullet points summarizing the article)

Writing style guidelines

• Lead with specifics, not abstractions — "12 lines of code" not "minimal boilerplate"
• Show, don't just tell — use code snippets as evidence
• Link generously — every mentioned file, class, or document should link to its repo URL
• Use subheadings within long sections for scannability
• One idea per paragraph — keep paragraphs to 3–5 sentences max

Step 4: Generate hashtags

Use the hashtag set from article-plan.md (or .config.md override). Total: 5–7 hashtags as a single copy-paste line.

Step 5: Validate

1. Word count is 1500–2500
2. All plan sections are present with correct headings
3. At least 3 code snippets are embedded
4. At least 5 repo links are included
5. No unsourced claims — every factual statement traces to claims.md
6. No forbidden phrases present
7. TL;DR block is present at the top

Step 6: Write output

Write 2 files.

File 1: content/article/draft.md

# [Article Title]

*[Subtitle]*

**Word count**: [N]
**Claims referenced**: [C-XXX, C-YYY, ...]
**Repo**: [repo URL]

---

## TL;DR

- [bullet 1]
- [bullet 2]
- [bullet 3]
- [bullet 4]

---

[Full article text with sections, code snippets, and links]

---

**Hashtags**: [copy-paste line]

File 2: content/article/hashtags.md

# Hashtags: Article

## Copy-paste block

[single line of 5–7 hashtags]

## Breakdown

- **Domain**: [hashtags grounded in language/platform/runtime, with rationale]
- **Topic-specific**: [hashtags grounded in this article's specific angle, with rationale]

Tip: paste with formatting

To paste the article into a rich-text editor (LinkedIn, Google Docs, Notion) with formatting preserved, render the markdown to HTML and pipe through the bundled helper:

pandoc content/article/draft.md -t html | "${CLAUDE_PLUGIN_ROOT}/scripts/pbcopy-html.swift"

For plain text with correct UTF-8 (handles bold/italic Unicode characters that macOS pbcopy mangles):

cat content/article/draft.md | "${CLAUDE_PLUGIN_ROOT}/scripts/pbcopy-text.swift"

Error handling

Condition	Response
Missing prerequisites	HARD BLOCK — see Prerequisites
Article plan has fewer than 6 sections	WARNING: proceed but note deviation
Code snippet source file unreadable	Skip snippet, add <!-- TODO: add code snippet from [file] --> placeholder
Word count outside 1500–2500 after drafting	Adjust: trim verbose sections or expand thin ones before final output