Content Pipeline

Content Pipeline

Orchestrate the full content creation workflow from research brief to published-ready content with social distribution pack.

Install

git clone https://github.com/thatrebeccarae/claude-marketing.git && cp -r claude-marketing/skills/content-pipeline ~/.claude/skills/

What This Skill Does

Chains up to 3 agents in sequence: 1. research-analyst — Generates a research brief on the topic 2. editor-in-chief — Reviews the draft against voice, structure, SEO, and argument quality 3. social-amplifier — Creates platform-specific distribution pack (LinkedIn, Twitter/X, email)

Users can enter at any stage or run the full pipeline.

How this differs from content-workflow: The content-workflow skill is a manual workflow guide with frameworks and checklists. This skill is the orchestration layer — it actually spawns subagents and manages the pipeline programmatically.

How to Use

/content-pipeline [topic or file path]

Full Pipeline (from scratch)

/content-pipeline "Why MCP is the new API"

Start from Existing Draft

/content-pipeline review ./drafts/article-draft.md

Social Pack Only

/content-pipeline distribute ./content/published-article.md

Workflow Stages

Stage 1: Research (optional — skip if draft exists)

• Spawn research-analyst agent with the topic
• Output: ./research-briefs/YYYY-MM-DD-topic-slug.md
• User reviews brief, decides whether to proceed

Stage 2: Editorial Review

• Spawn editor-in-chief agent pointing at the draft/brief
• Output: [draft-path]-review.md saved alongside the source
• Agent provides: voice assessment, structure feedback, SEO evaluation, line-level notes, title variants
• User decides: publish as-is, revise, or iterate

Stage 3: Social Distribution

• Spawn social-amplifier agent pointing at the final content
• Output: [content-path]-social-pack.md saved alongside the source
• Generates: LinkedIn posts (2 variants), Twitter/X thread, email subject lines, pull quotes, hashtags

Configuration

Customizing Agents

The default agent chain uses Claude Code subagents with the Agent tool. To customize:

Pipeline Stage    | Default Agent Type    | Override With
Research          | research-analyst      | Any research/analysis agent
Editorial         | editor-in-chief       | Any editorial review agent
Distribution      | social-amplifier      | Any social content agent

Output Paths

Configure where each stage saves its output:

Publishing Cadence

Adapt the pipeline to your publishing schedule. Example cadence:

• Research: Monday/Tuesday
• Draft: Tuesday/Wednesday
• Review: Wednesday/Thursday
• Publish: Thursday/Friday

Pipeline State Management

The pipeline tracks which stages are complete via file existence:

• Research brief exists → Stage 1 complete
• Review file exists → Stage 2 complete
• Social pack exists → Stage 3 complete

Resume from any stage by pointing to existing files.

Integration with Other Skills

This skill composes well with:

• seo-content-writer — Run during Stage 2 for deep SEO optimization
• content-creator — Use for brand voice analysis before editorial review
• research-digest — Alternative to the built-in research stage for RSS-based research

Quick Commands

• /content-pipeline — Start full pipeline with topic prompt
• /content-pipeline review [path] — Jump to editorial review
• /content-pipeline distribute [path] — Jump to social distribution
• /content-pipeline status — Show drafts in progress

Content Pipeline — Examples

Example 1: Full Pipeline from Topic

Prompt

/content-pipeline "Why email deliverability matters more than open rates"

What happens

1. Stage 1 — Research: Spawns research-analyst agent

• Searches for recent data on email deliverability vs open rate metrics
• Finds industry reports, expert opinions, and case studies
• Outputs: ./research-briefs/2026-03-18-email-deliverability-vs-open-rates.md
• User reviews brief and approves direction

2. Stage 2 — Editorial Review: Spawns editor-in-chief agent

• Reviews the draft for voice consistency, argument strength, and SEO
• Checks keyword integration for "email deliverability"
• Outputs: ./content/2026/2026-03-18-SS-email-deliverability.md-review.md
• Decision: REVISE (suggests stronger opening hook and more data points)

3. Stage 3 — Social Distribution: After revisions, spawns social-amplifier

• Generates 2 LinkedIn post variants (story-led and data-led)
• Creates 6-tweet Twitter thread
• Produces 3 email subject line options
• Outputs: ./content/2026/2026-03-18-SS-email-deliverability.md-social-pack.md

---

Example 2: Review Existing Draft

Prompt

/content-pipeline review ./drafts/email-deliverability.md

What happens

Skips Stage 1 (research) and goes directly to editorial review:

1. Stage 2 — Editorial Review: Reads the draft at the specified path

• Assesses voice: "Conversational but authoritative — matches target"
• Checks structure: "Missing a counter-argument section"
• SEO check: "Primary keyword appears in title and H2s, missing from meta description"
• Provides 5 line-level suggestions
• Offers 3 alternative titles
• Decision: PASS (with minor revisions noted)

2. Stage 3 — Social Distribution: Proceeds automatically after PASS

• Generates full social distribution pack

---

Example 3: Generate Social Pack from Published Article

Prompt

/content-pipeline distribute ./content/2026/2026-03-15-SS-mcp-future-of-apis.md

What happens

Skips Stages 1 and 2, goes directly to distribution:

1. Stage 3 — Social Distribution: Reads the published article

• LinkedIn Variant A (story-led): Opens with a personal anecdote about API integration pain
• LinkedIn Variant B (data-led): Opens with "73% of developers say API integration is their biggest time sink"
• Twitter thread: 7 tweets distilling the key argument
• Email subjects: "The API is dead. Long live MCP.", "Why I stopped building API wrappers", "MCP changed how I think about integrations"
• Pull quotes: 2 quotable passages for social sharing
• Hashtags: LinkedIn set (#MCP #APIs #DevTools) and Twitter set (#buildinpublic #devtools)

Content Pipeline — Reference

Agent Spawning Patterns

Research Agent

Spawn a research-analyst agent with:
- Topic: [user-provided topic]
- Output format: structured research brief with sections for key findings, data points, expert perspectives, counter-arguments, and content angles
- Output path: ./research-briefs/YYYY-MM-DD-topic-slug.md

Editorial Agent

Spawn an editor-in-chief agent with:
- Input: [path to draft or brief]
- Review criteria: voice consistency, argument structure, SEO optimization, factual accuracy, readability
- Output format: review document with inline comments, overall assessment, and revision suggestions
- Output path: [input-path]-review.md

Social Amplification Agent

Spawn a social-amplifier agent with:
- Input: [path to final content]
- Output format: platform-specific distribution pack
- Output path: [input-path]-social-pack.md

Output Format Specifications

Research Brief Format

# Research Brief: [Topic]

**Date:** YYYY-MM-DD
**Topic:** [Topic description]
**Sources:** [Number] sources consulted

## Key Findings
1. [Finding with source citation]
2. [Finding with source citation]

## Data Points
- [Statistic or metric with source]

## Expert Perspectives
- [Expert name/org]: [Key quote or position]

## Counter-Arguments
- [Opposing view and its strength]

## Content Angles
1. [Angle 1]: [Why it works]
2. [Angle 2]: [Why it works]

## Recommended Direction
[Which angle to pursue and why]

Editorial Review Format

# Editorial Review: [Title]

**Reviewed:** YYYY-MM-DD
**Source:** [Path to draft]
**Decision:** PASS | REVISE | ESCALATE

## Voice Assessment
- **Consistency:** [Score 1-5]
- **Notes:** [Specific feedback]

## Structure
- **Flow:** [Assessment]
- **Argument strength:** [Assessment]

## SEO Evaluation
- **Primary keyword:** [Keyword] — [Present/Missing]
- **Meta description:** [Assessment]
- **Heading hierarchy:** [Assessment]

## Line-Level Notes
- Line [N]: [Specific suggestion]

## Title Variants
1. [Alternative title 1]
2. [Alternative title 2]
3. [Alternative title 3]

Social Distribution Pack Format

# Social Distribution Pack: [Title]

**Generated:** YYYY-MM-DD
**Source:** [Path to content]

## LinkedIn Post (Variant A — Story-led)
[Full post text, 1200-1500 chars]

## LinkedIn Post (Variant B — Data-led)
[Full post text, 1200-1500 chars]

## Twitter/X Thread (5-7 tweets)
1/ [Tweet 1 — hook]
2/ [Tweet 2]
...

## Email Subject Lines
1. [Subject line 1]
2. [Subject line 2]
3. [Subject line 3]

## Pull Quotes
- "[Quote 1]"
- "[Quote 2]"

## Hashtags
[Platform-specific hashtag sets]

Error Handling

Stage Failures

Retry Patterns

• Max 2 retries per stage before asking user for direction
• Each retry should include feedback from the failed attempt
• Never silently skip a stage — always inform the user