contentforge

Orchestrates a 10-phase subagent pipeline producing fact-checked, SEO-optimized articles as .docx files.

1|Updated Aug 18, 2026
One-click install
npx skills add https://github.com/teachskillofskills-ai/ContentForge-techshu --skill contentforge-teachskillofskills-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: contentforge
Source: https://github.com/teachskillofskills-ai/ContentForge-techshu/tree/main/skills/contentforge
Command: npx skills add https://github.com/teachskillofskills-ai/ContentForge-techshu --skill contentforge-teachskillofskills-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Producing long-form marketing content that is fact-checked, on-brand, SEO-optimized, and free of detectable AI-writing patterns normally requires many manual steps and editorial passes, and single-shot AI generation tends to hallucinate statistics and skip verification. ## Core Features & Use Cases - 10-phase agent pipeline: Dedicated subagents handle research, fact-checking, drafting, visual annotation, scientific validation, structuring, SEO/GEO optimization, humanizing (43-pattern catalog), 5-dimension review scoring, and .docx output, each behind an orchestrator-verified quality gate. - Checkpointed, resumable runs: Every phase artifact is checkpointed to a run directory, with loop limits, a run auditor that re-derives every gate, and cross-session resume support. - Express and bring-your-own-draft modes: Run a reduced lane using user-supplied sources, or preserve an author's own sentences verbatim while building the article around them. - Use Case: A three-person marketing team needs 30 cited, brand-compliant articles in a quarter; running the pipeline per topic yields a publication-ready .docx with scorecard appendices in 30-60 minutes per piece. ## Quick Start Ask the assistant to run /contentforge:contentforge with your topic, brand slug, content type, and target keyword to produce a fact-checked publication-ready article.

Frequently Asked Questions about contentforge

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a fact-checked SEO article with an AI pipeline?

Run the contentforge pipeline with a topic, brand slug, content type, and primary keyword. It executes research, fact-checking, drafting, validation, SEO, humanizing, and review phases, each verified by a quality gate, and delivers a .docx with scorecard appendices.

How does the fact-checking phase prevent hallucinated statistics?

Phase 2 verifies at least 80% of claims, requires all cited URLs to be live, and tolerates at most 3 unverified claims. Phase 4 then validates that every claim traces to a cited source, with a fix ledger tracking every correction.

Can I use my own draft or sources instead of full research?

Yes. The --source-draft flag preserves your sentences verbatim while building the article around them, and express mode catalogs only your provided sources while keeping the fact-check, validation, and review gates unchanged.

Does the pipeline work on platforms without subagent support?

Yes. A portable execution lane runs phases sequentially in a single context by reading each agent file as instructions, producing the same artifacts, gates, and checkpoints as subagent-dispatched runs.

What happens if a pipeline run is interrupted mid-phase?

Every gate-passed phase is checkpointed to a run directory with a run.json manifest. Running the resume command reloads only the artifacts the next phase needs and continues, with orphaned-artifact detection preventing duplicate work.

When should I not use the full content pipeline?

Skip the full pipeline when you already have verified sources and need speed; express mode drops research, visuals, and SEO phases while keeping verification gates. For multiple pieces, use the batch-process skill instead of repeated single runs.