Content

Content Skill

Six modes: Calendar (editorial pipeline), Marketing (SEO, campaigns, email, competitive), Publish (outline, pre-publish, SEO, audits, WordPress), Social-X (X/Twitter posts and threads), Social-Bluesky (public feed reading), Social-Reddit (moderation). Classify the request and follow the matching section.

Mode Selection

Calendar Mode

Manage editorial content through 6 pipeline stages: Ideas, Outlined, Drafted, Editing, Ready, Published. All state lives in a single content-calendar.md file.

Workflow

1. READ -- Load and parse content-calendar.md. Validate all sections exist. Do not trust memory; always re-read the actual file.
2. EXECUTE -- Perform the requested operation:
   • View: count per stage, upcoming scheduled (14 days), in-progress, recent publications (30 days). Flag stuck content (14+ days in one stage).
   • Add idea: check for duplicate titles (case-insensitive), append to Ideas.
   • Move: forward only through defined sequence (Ideas -> Outlined -> Drafted -> Editing -> Ready -> Published). Record YYYY-MM-DD timestamp at each transition.
   • Schedule: set date on Ready items. Date must be today or future.
   • Archive: move Published entries older than current month to Historical sections.
3. WRITE -- Persist changes, re-read file to verify.

Calendar Sub-modes

Marketing Mode

Seven sub-modes. Always load references/llm-marketing-failure-modes.md alongside mode-specific references for detection heuristics against generic copy, keyword stuffing, fabricated metrics.

SEO Audit Quick Steps

1. Gather: URL/domain, audit type, target keywords, competitors (identify 2-3 via web search if not provided).
2. Keyword research: classify by intent (informational, navigational, commercial, transactional). Assess difficulty and opportunity.
3. On-page audit: title tags, meta descriptions, H1/H2, keyword usage, internal linking, image alt text, URL structure.
4. Technical: page speed, mobile, structured data, crawlability, HTTPS, Core Web Vitals.
5. Content gaps: competitor topic coverage, freshness, thin content, funnel gaps.
6. Output: executive summary, keyword table (15-25), issues table, content gap recommendations, technical checklist, competitor matrix, prioritized action plan.

Publish Mode

Eight sub-modes covering the blog publishing pipeline.

Outline Quick Steps

1. Gather: topic, target audience, desired format, word count range, SEO target keyword.
2. Research: check existing coverage, identify gaps, gather supporting data.
3. Structure: working title, hook, thesis, 3-7 sections with sub-points, conclusion CTA.
4. Validate: logical flow, sufficient depth per section, SEO keyword placement plan.

Pre-publish Quick Steps

1. Check: title, date, slug, description, tags, categories, featured image, author, draft=false.
2. Content: spell check, broken internal links, image paths, code blocks render, meta descriptions.
3. SEO: title tag (<60 chars), meta description (120-160 chars), keyword in H1 and first paragraph.

Social-X Mode

OAuth-authenticated X/Twitter posting via x-api-poster.py. 4-phase pipeline with confirmation gate.

Workflow

1. VALIDATE -- Check credentials: python3 $HOME/.claude/scripts/x-api-poster.py post --dry-run --text "ping". Requires X_API_KEY, X_API_SECRET, X_ACCESS_TOKEN, X_ACCESS_SECRET, X_BEARER_TOKEN. Read-only ops need only X_BEARER_TOKEN.
2. CONFIRM -- Show user the exact text. For threads: python3 $HOME/.claude/scripts/x-api-poster.py thread --dry-run --texts "part 1" "part 2". Enforce 280-char limit per tweet. Wait for explicit user confirmation.
3. POST -- Execute with --confirmed flag: python3 $HOME/.claude/scripts/x-api-poster.py post --confirmed --text "text". For media: --media /path/to/image.png.
4. VERIFY -- Report tweet URL and rate limit status.

Social-Bluesky Mode

Read public Bluesky feeds via AT Protocol. No auth required.

Commands

bashCopy code
# Fetch recent posts
python3 ~/.claude/scripts/bluesky_reader.py feed --handle HANDLE --limit 20

# Search posts
python3 ~/.claude/scripts/bluesky_reader.py search --handle HANDLE --query "terms"

# JSON output for pipelines
python3 ~/.claude/scripts/bluesky_reader.py feed --handle HANDLE --json

Endpoint: https://public.api.bsky.app/xrpc/app.bsky.feed.getAuthorFeed. Limit: 1-100 posts. Search is local keyword filter (all words must appear, case-insensitive).

Social-Reddit Mode

Reddit moderation via PRAW. Three modes:

Workflow

1. FETCH -- python3 scripts/reddit-moderate/reddit-mod.py queue --json --limit 25 | python3 scripts/reddit-moderate/reddit-mod.py classify. Loads subreddit context from reddit-data/{subreddit}/.
2. CLASSIFY -- Fill classification, confidence, and reasoning for each item using the assembled prompt.
3. PRESENT -- Group by confidence tier (high/medium/low). Show: title, author, report reason, classification, action, reasoning.
4. ACT -- Execute confirmed actions: python3 scripts/reddit-moderate/reddit-mod.py action --item-id ID --action approve|remove|spam.

Deep References

Load when the task needs detailed patterns, specifications, or examples.

Calendar

Marketing

Publish

Social

Scripts