ContentForge — Enterprise Content Production

Transform a content requirement into a publication-ready, fact-checked, brand-compliant, SEO-optimized piece in 20-30 minutes through a 10-phase autonomous agent pipeline with three-layer fact verification and zero hallucinations.

Context efficiency

Pipeline phase. Grep before Read for references/, humanization-patterns.json, brand voice profiles. Pass earlier-phase outputs by path + line range, not by reloading. On /contentforge:resume, load only the failed phase's state.

Execution Protocol (CRITICAL — read first)

This skill orchestrates 11 distinct phases. Each phase MUST be executed by invoking its dedicated subagent via the `Task` tool — DO NOT generate the deliverable yourself in a single inference pass. A single-pass generation skips the quality gates, fact-checking layers, humanizer 29-pattern catalog, and reviewer scoring that define ContentForge.

Required Per-Phase Workflow

For every one of the 11 phases below:

1. Call `Bash` to mark phase start:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/pipeline-tracker.py --action phase-start --brand <slug> --phase <N>

1. Call `Task` with the phase's `subagent_type` and pass the previous phase's output as context.
2. Call `Bash` to mark phase end with the output word count:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/pipeline-tracker.py --action phase-end --brand <slug> --phase <N> --content-words <count>

1. Emit the audit line (so users can see real-time progress):

   [PHASE-AUDIT] phase=<N> name=<name> status=<PASS|FAIL> output_summary="<one line>" gate=<PASS|FAIL>

1. Check the quality gate. If gate=FAIL, loop back to the phase the reviewer flagged (max 5 total loops; on 5th loop, escalate to human review and halt).

Phase → Subagent Mapping

Final Output Requirements

After Phase 8 completes, the output-manager subagent must produce a Microsoft Word `.docx` file by calling:

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/generate-docx.py \
    --content <article.md> \
    --output <local-path>.docx \
    --reports <reports.json> \
    --brand "<brand>" \
    --content-type <type>

The .docx must contain: title page, full article body, sources/citations, Appendix A (SEO Scorecard), Appendix B (Quality Scorecard), Appendix C (Production Details).

If the brand has Google Drive configured (tracking.backend == "google"), upload the .docx via drive-uploader.py. Otherwise save locally to ~/.claude-marketing/<brand>/output/<type>/<YYYY-MM-DD>/<slug>.docx and tell the user the local path.

Why This Matters

Skipping the Task-tool orchestration means: no real fact-checking, no real humanizer (29-pattern AI removal won't fire), no real reviewer scoring — the pipeline becomes single-pass content generation labeled with fake phase names. The audit trail (pipeline-run.json, [PHASE-AUDIT] lines, real reviewer score) is the proof of execution. If those artifacts don't exist after a run, the pipeline didn't actually run.

When to Use

Use /contentforge when you need:

• Single high-quality content piece (article, blog, whitepaper, FAQ, research paper)
• Research-backed content with verified citations
• Brand-compliant content for regulated industries (Pharma, BFSI, Healthcare, Legal)
• SEO-optimized content with keyword targeting and meta tags
• Natural-sounding content with AI patterns removed (Phase 6.5 Humanizer)

For multiple pieces in parallel, use /batch-process instead (4-5x faster).

What This Command Does

Runs your content through 10 specialized agents with quality gates at each phase:

1. Research Agent — SERP analysis, source mining, competitive analysis, structured outline
2. Fact Checker — URL verification, claim validation, confidence scoring
3. Content Drafter — First draft with brand voice, SME calibration via industry knowledge packs
4. Visual Asset Annotator — Chart generation from verified stats, visual markers, asset manifest
5. Scientific Validator — Hallucination detection, domain-specific validation, logic validation
6. Structurer & Proofreader — Grammar/spelling correction, readability optimization, brand compliance
7. SEO/GEO Optimizer — Keyword optimization, meta tag generation, internal linking markers
8. Humanizer — AI pattern removal, sentence variety (burstiness), brand personality
9. Reviewer — 5-dimension quality scoring (Content Quality 30%, Citation Integrity 25%, Brand Compliance 20%, SEO Performance 15%, Readability 10%)
10. Output Manager — .docx with embedded charts, internal links, Google Drive upload

Quality Gates: If any phase fails, the pipeline loops back with feedback (max 5 total loops before human escalation).

Required Inputs

Minimum Required:

• Topic — What the content is about (e.g., "AI in Healthcare", "remote work productivity")
• Content Type — article, blog, whitepaper, faq, research_paper
• Brand — Which brand profile to use (create with /contentforge:style-guide if new brand)

Pre-Flight Validation: After gathering inputs, the system validates your brand profile for completeness (voice, guardrails, audience, industry pack). For regulated industries (pharma, BFSI, healthcare, legal), guardrails are required — the system will warn if they're empty and ask whether to proceed or update the profile first.

Optional:

• Target Audience — Who this content is for (e.g., "Healthcare CIOs")
• Word Count — Target length (defaults to content type standard)
• Primary Keyword — Main SEO keyword to optimize for
• Tone — Overrides brand default (authoritative, conversational, technical, witty)

How to Use

Interactive Mode (Recommended for First-Time Users)

/contentforge

Prompts you for:

1. Topic (the subject — NOT the final title)
2. Content Type (select from 5 options)
3. Brand (select from existing profiles)
4. Target Audience
5. Word Count (or use default)
6. Primary Keyword

Then generates 4-5 title options (different angles: benefit-driven, how-to, data-driven, question-based, contrarian). You select, modify, or provide your own title. Pipeline starts only after title confirmation.

Quick Mode (Topic Provided)

/contentforge "AI in Healthcare" --type=article --brand=AcmeMed --audience="Healthcare CIOs" --keyword="AI healthcare 2026"

Even in quick mode, the system generates title options and asks you to select before starting Phase 1. The topic you provide is the subject — the final title is always a user decision.

Use Existing Google Sheet Requirement

/contentforge --sheet-url=https://docs.google.com/spreadsheets/d/ABC123 --row=5

Reads requirement from Row 5 of the sheet.

What Happens

Title Curation (1-2 minutes) — MANDATORY

Before the pipeline starts, the system generates 4-5 SEO-optimized title options using the topic, content type, brand voice, audience, and primary keyword. Each title takes a different angle:

• Benefit-driven — leads with reader value
• How-to / Tactical — actionable, instructional
• Data-driven / Stat-led — opens with a number or trend
• Question-based / Curiosity — provokes the reader
• Contrarian / Unexpected — challenges convention

You select, modify, or provide your own title. The confirmed title becomes the anchor for the entire pipeline — research, outline, SEO, and final output all flow from it.

Do NOT skip this step or auto-select a title. The title shapes the entire content piece.

Phase 1: Research (3-5 minutes)

• Uses the confirmed title as the anchor for all research
• Performs SERP analysis for the topic and title angle
• Mines 10-15 authoritative sources
• Analyzes competitor content
• Generates structured outline aligned with the confirmed title
• Quality Gate: Must have 5+ live sources, differentiated angle

Phase 2: Fact Checking (2-3 minutes)

• Verifies all URLs are accessible (no 404s)
• Validates claims against sources
• Assigns confidence scores (strongly verified, partially verified, weakly verified)
• Flags any unverifiable claims
• Quality Gate: 80%+ verified claims, zero flagged items, all URLs live

Phase 3: Content Drafting (5-7 minutes)

• Generates first draft with brand voice
• Includes inline citations (APA format)
• Targets word count ±10%
• Maintains min 1 citation per 300 words
• Quality Gate: Word count ±10%, all outline sections covered, citation density met

Phase 4: Scientific Validation (2-3 minutes)

• Scans for hallucinations (fabricated statistics, made-up studies)
• Ensures all claims are traceable to sources
• Validates logical consistency
• Quality Gate: Zero hallucinations, all claims traceable
• If fails: Loops back to Phase 3 with specific claims to fix (max 2 loops)

Phase 5: Structuring & Proofreading (2-3 minutes)

• Corrects grammar and spelling (100% accuracy)
• Optimizes readability for content type (Grade 8-16 depending on type)
• Enforces brand terminology and style guide
• Quality Gate: Zero grammar errors, readability on target, 100% brand compliant

Phase 6: SEO/GEO Optimization (2-3 minutes)

• Optimizes keyword density (target: 1.5-2.5%)
• Places keywords in critical positions (title, H2s, first paragraph, conclusion)
• Generates meta title, meta description, URL slug
• Prepares content for AI answer engines (ChatGPT, Perplexity, Gemini)
• Quality Gate: Keyword density 1.5-2.5%, all critical placements hit, meta tags optimized

Phase 6.5: Humanizer ⭐ (1-2 minutes)

• Removes AI telltale phrases (20+ patterns: "delve", "leverage", "it's important to note")
• Increases sentence variety (burstiness ≥0.7 for natural human rhythm)
• Injects brand personality (authoritative, witty, warm, data-driven)
• Validates SEO preservation (keyword density unchanged ±2 occurrences)
• Quality Gate: AI patterns removed, burstiness ≥0.7, SEO preserved

Phase 7: Reviewer (2-3 minutes)

• Scores content across 5 dimensions:
• Content Quality (30%): Depth, originality, value, clarity
• Citation Integrity (25%): Accuracy, relevance, authority, freshness
• Brand Compliance (20%): Voice, terminology, guardrails, style
• SEO Performance (15%): Keyword optimization, meta tags, structure
• Readability (10%): Grade level, sentence variety, flow
• Calculates composite score (1-10, needs ≥7.0 to pass)
• Quality Gate: Score ≥7.0, all dimensions pass, zero critical violations
• If <7.0: Loops back to failing phase with specific feedback (max 2 loops)

Phase 8: Output Manager (1-2 minutes)

• Generates .docx file with proper formatting
• Uploads to Google Drive (ContentForge Output/[Brand]/[Title]_v1.0.docx)
• Updates tracking sheet with:
• Requirement ID, Title, Brand, Type, Word Count
• Quality Score (breakdown by dimension)
• Processing Time, Completion Timestamp
• Output URL (Drive link)

Output Example

Every pipeline run ends with a Completion Card showing scores, stats, timing, and delivery status. This card is mandatory — it's shown in the conversation AND added as an appendix in the .docx file.

Example Completion Card:

CONTENTFORGE — COMPLETION CARD

Content:  "AI in Healthcare: 2026 Trends" | AcmeMed | Article | ✅ APPROVED

Quality Score: 9.2/10 (Grade A+)
  Content Quality:    9.5/10 (30%) ✅
  Citation Integrity: 9.0/10 (25%) ✅
  Brand Compliance:   9.5/10 (20%) ✅
  SEO Performance:    8.8/10 (15%) ✅
  Readability:        9.0/10 (10%) ✅

Content Stats:
  Words: 1,947 (target 1,500-2,000) ✅ | Citations: 14 sources ✅
  Keyword: "AI healthcare 2026" at 2.1% ✅ | Readability: Grade 11.2 ✅
  Burstiness: 0.78 ✅ | AI Patterns: 0 remaining ✅ | Hallucinations: 0 ✅

SEO Package:
  Meta Title: "AI in Healthcare: 2026 Trends..." (58 chars) ✅
  Meta Description: "Discover how AI is transforming..." (152 chars) ✅
  Internal Links: 4 applied | Feature Image: generated (user-approved)

Pipeline: 24 min total | 0 loops | Guardrails: verified
  Research 4m | Fact-Check 3m | Draft 6m | Visuals 2m | Validate 2m
  Structure 2m | SEO 2m | Humanize 1m | Review 2m | Output 1m

Delivery:
  .docx: ✅ Generated
  Google Drive: ✅ ContentForge Output/AcmeMed/AI-in-Healthcare-2026-Trends_v1.0.docx
  Tracking: ✅ Row 5 updated

Next: /contentforge:publish | /contentforge:social-adapt | /contentforge:translate | /contentforge:variants

Content Types & Specifications

Brand Profile Setup

Before using ContentForge, create a brand profile:

/contentforge:style-guide

Provide your brand name, industry, voice guidelines (or share existing documents/URLs), and ContentForge generates the profile JSON automatically.

Alternatively, copy config/brand-registry-template.json and fill in manually.

Brand Profile Includes:

• Voice & Tone (authoritative, conversational, technical, witty)
• Terminology (approved terms, banned phrases)
• Style Guide (formatting preferences, citation style)
• Guardrails (topics to avoid, compliance requirements)
• Industry Context (Pharma, BFSI, Healthcare, Legal)
• Personality Profile (new in v3.0: authoritative, conversational, technical, witty)

Brand profiles are cached (SHA256 hash) for 95% time savings on repeat runs.

See the User Guide for detailed setup instructions.

Quality Assurance

Three-Layer Fact Verification

1. Phase 2 (Fact Checker): URL verification, claim validation
2. Phase 4 (Scientific Validator): Hallucination detection
3. Phase 7 (Reviewer): Final citation integrity scoring

Result: Zero hallucinations in production testing

Feedback Loop Management

• Phase 4 → Phase 3: Max 2 loops (hallucination fixes)
• Phase 7 → Any Phase: Max 2 loops (quality improvements)
• Total Loop Limit: 5 iterations before human escalation

Human Review Escalation

Content is flagged for human review if:

• Quality score <5.0/10 after max loops
• Critical brand violations detected
• Excessive loops without improvement
• User explicitly requests review

Flagged content is NEVER auto-published.

Performance Metrics

Typical Processing Times:

• Blog (1,200 words): 15-20 minutes
• Article (1,800 words): 22-28 minutes
• Whitepaper (3,500 words): 30-40 minutes

Quality Scores (Avg across 200+ pieces in beta):

• Overall: 8.7/10
• Content Quality: 8.9/10
• Citation Integrity: 8.5/10
• Brand Compliance: 9.2/10
• SEO Performance: 8.6/10
• Readability: 8.8/10

Accuracy:

• Factual Accuracy: 100%
• Citation Accuracy: 95%+
• Brand Compliance: 100%
• Hallucinations: 0

Integration with Other Skills

Before ContentForge:

• /contentforge:style-guide — Create brand profile if new brand
• /contentforge:brief — Generate research-backed content brief with keyword analysis

Instead of ContentForge (for scale):

• /batch-process — Process 10-50+ pieces in parallel (4-5x faster)

After ContentForge:

• /content-refresh — Update content 6-12 months later with fresh data
• /contentforge:variants — Create A/B test headline/hook/CTA variations
• /contentforge:publish — Publish to Webflow or WordPress via MCP
• /contentforge:social-adapt — Transform article into LinkedIn, Twitter/X, Instagram, Facebook, Threads posts
• /contentforge:translate — Translate preserving brand voice (15+ languages)
• /contentforge:video-script — Generate timestamped video scripts from the article
• /contentforge:analytics — Record quality scores for trend tracking

Requirements

MCP Integrations (Optional)

• Google Sheets — Requirement intake for batch processing, quality tracking
• Google Drive — Brand knowledge vault, output .docx storage
• Webflow/WordPress — Direct CMS publishing via /contentforge:publish

Run /contentforge:integrations to check your connector status. Run /contentforge:connect <name> for setup guides.

Environment

• Claude Code or Cowork (latest version)
• Internet connection (for Phase 1 web research)

Troubleshooting

"Brand profile not found"

When: You run /contentforge with a brand that doesn't have a profile yet.

Fix:

1. Create a brand profile (recommended, 5 min):

   /contentforge:style-guide

Answer 3 questions (name, tone, industry) and you're ready.

1. Or specify a different brand:

   /contentforge "your topic" --brand=ExistingBrand

"Quality score <5.0, flagged for review"

When: Content didn't meet the minimum quality threshold after all feedback loops.

Common causes and fixes:

• Topic too vague → Be more specific: "AI in healthcare" → "AI diagnostic tools for rural hospitals in 2026"
• Sources behind paywalls → Provide accessible reference URLs with --sources=
• Brand profile incomplete → Run /contentforge:style-guide --update [brand] to add guardrails and terminology
• Niche topic with few sources → Consider a broader angle or provide your own source URLs

"Max loops exceeded (5 iterations)"

When: The pipeline kept trying to improve content but couldn't reach the quality threshold.

What happened: Phase 7 (Reviewer) scored the content below 7.0 multiple times and looped back to fix it, but improvements plateaued.

Fix:

1. Check which dimension scored lowest (Content Quality? Citations? Brand Compliance?)
2. If Content Quality is low → topic needs more depth or the angle is too broad
3. If Citation Integrity is low → sources are weak or behind paywalls
4. If Brand Compliance is low → brand profile may be incomplete
5. Re-run with adjustments: more specific topic, better keywords, or updated brand profile

"Processing time >45 min for article"

When: The pipeline is taking longer than expected.

This is usually normal — API rate limits or network latency cause delays. ContentForge auto-retries with backoff.

If it persists beyond 60 min:

1. Check internet connection
2. Run /contentforge:integrations to verify MCP servers are responding
3. Try a simpler topic to isolate the issue
4. Large whitepapers (5000+ words) can legitimately take 45-75 min

"Guardrails empty — compliance skipped"

When: Your brand profile doesn't have prohibited claims or required disclaimers defined.

Impact: Phase 5 (Brand Compliance) will report "SKIPPED" instead of actually checking content. Phase 7 applies a -1.0 penalty to Brand Compliance score.

Fix:

/contentforge:style-guide --update [brand]

Add at minimum: 3-5 prohibited claims, any required legal disclaimers, and industry-specific restrictions.

For regulated industries (pharma, BFSI, healthcare, legal): This is critical. Empty guardrails mean no compliance verification.

Pipeline phase explanations

During content production, you'll see updates as each phase completes:

If a phase loops back: The system shows which phase failed, why, and what it's fixing. Loops are automatic — you don't need to do anything unless it escalates to human review.

Example Workflow

Scenario: Create 1 thought leadership article for AcmeMed brand

Step 1: Create Brand Profile (One-Time Setup)

/contentforge:style-guide

Provide: Brand name (AcmeMed), Industry (Healthcare), Voice (Authoritative), Tone (Professional), Terminology, Guardrails

Step 2: Start Content Production

/contentforge "AI-Powered Diagnostics in Precision Medicine" --type=article --brand=AcmeMed --audience="Healthcare Executives" --keyword="AI diagnostics precision medicine"

Step 3: Select Title (1-2 minutes)

ContentForge generates 4-5 title options:

1. "AI-Powered Diagnostics: The Future of Precision Medicine in 2026"
2. "How AI Diagnostics Are Transforming Precision Medicine for Healthcare Leaders"
3. "5 AI Diagnostic Breakthroughs Reshaping Precision Medicine Right Now"
4. "The Executive's Guide to AI-Powered Precision Medicine Diagnostics"
5. "Why AI Diagnostics in Precision Medicine Are Finally Delivering on the Promise"

You select Option 1 → Pipeline starts with that title as the anchor.

Step 4: Review Output (24 minutes later)

• Quality Score: 9.1/10 ✅
• Word Count: 1,922 ✅
• Citations: 12 sources ✅
• SEO: Keyword density 2.3% ✅

Step 5: Publish

/contentforge:publish --platform=webflow

Total Time: 25 minutes (setup once, then 20-30 min per piece)

Limitations

• Sequential processing (for parallel, use /batch-process)
• 20-30 min per piece (cannot be rushed without compromising quality)
• Best with brand profile — works without one but uses generic defaults

Related Skills

• [/batch-process](../batch-process/SKILL.md) — Process 10-50+ pieces in parallel (4-5x faster)
• [/content-refresh](../content-refresh/SKILL.md) — Update old content with fresh data
• [/contentforge:variants](../cf-variants/SKILL.md) — A/B test headline/hook/CTA variations
• [/contentforge:analytics](../cf-analytics/SKILL.md) — Track quality scores and performance
• [/contentforge:social-adapt](../cf-social-adapt/SKILL.md) — Transform article into social media posts
• [/contentforge:publish](../cf-publish/SKILL.md) — Publish to Webflow/WordPress
• [/contentforge:translate](../cf-translate/SKILL.md) — Translate preserving brand voice
• [/contentforge:brief](../cf-brief/SKILL.md) — Generate research-backed content briefs

<!-- Version, agent count, and asset counts are pulled live by /contentforge:cf-help from scripts/plugin-metadata.py. Do not bake version strings into skill bodies -- they drift out of sync every release. The canonical source of truth is .claude-plugin/plugin.json + the agents/ + skills/ directories. -->

Pipeline: 10 phases plus Step 0.5 (Title Curation). Phase agents are defined in agents/*.md and enumerated by scripts/plugin-metadata.py --section pipeline. Post-pipeline agents include Batch Orchestrator, Social Adapter, and Translator.

Processing Time: 20-30 minutes avg for an article on Opus 4.7. Slower on smaller models and for whitepaper / research-paper content types.

Quality target: composite Reviewer score ≥7.0 to pass; max 2 loops per phase; three-layer verification (Fact Checker → Scientific Validator → Reviewer) for zero-hallucination production output.