content-factory

Multi-agent content production system. One piece of source content becomes many formats — social posts, email, scripts, headlines, and more. Five specialized...

Security Scan

The skill's requirements and instructions are coherent with its stated purpose (multi-agent content production); it is an instruction-only skill that doesn't request extra credentials or installs, though you should be mindful about optional network usage and voice-cloning misuse.

Purpose & Capability

Name/description (content production, multi-agent personas) match the provided files and templates. No unexpected binaries, env vars, or config paths are requested. All declared behaviors (reading source content, writing drafts) are appropriate for the stated purpose.

Instruction Scope

SKILL.md and prompt files describe only content-generation tasks and expected I/O (read source content, write outputs). A couple of items merit attention: the 'voice-cloner' prompt explicitly trains/writes in another author's voice (which can enable stylistic impersonation) and the research pipeline expects source URLs/excerpts (which implies fetching/using external sources if network access is enabled). These are consistent with the purpose but are higher-risk uses of content-generation features and should be used with policy safeguards and user consent.

Install Mechanism

No install spec and no code files — instruction-only. This is low-risk from an installation/execution perspective because nothing is downloaded or installed by the skill itself.

Credentials

The skill declares no required environment variables, credentials, or config paths — proportional and minimal. It does note 'sub-agent spawning' as optional; if you enable that or grant network access, external API keys or cloud credentials could become relevant in practice, so only enable those capabilities when needed and understand what endpoints will be called.

Persistence & Privilege

always:false and the skill does not request permanent presence or modify other skills. Default autonomous invocation remains possible, which is platform-normal and not itself a negative here.

Assessment

This skill appears internally consistent for generating and remixing content; it doesn't ask for secrets or install anything. Before installing: (1) Decide whether to enable network/sub-agent spawning — if you enable it, the agent could fetch URLs or call cloud APIs, so only enable with appropriate access controls. (2) Avoid feeding private or sensitive data (personal PII, confidential documents, or proprietary source material) into the skill. (3) Be cautious with the voice-cloner prompt — it can reproduce writing styles and could be misused for impersonation; only use it with samples/consent you are allowed to copy. (4) Review generated content for factual accuracy, copyright, and any policy/theological constraints referenced in the prompts. If you want, restrict the skill's network capability or inspect agent logs the first few runs to confirm behavior matches your expectations.

Content Factory — Multi-Agent Content Production System

One source → many formats. One system → consistent brand voice.

---

What This Is

Content Factory is a structured system for content production. Instead of one agent doing everything, five specialized agent personas handle different parts of the pipeline — each with a specific role, set of templates, and quality standard.

Load this skill when:

• User wants to create content from scratch, a topic, or a research dump
• User has one piece of content and wants it adapted to multiple platforms
• User needs consistent brand voice across formats
• User wants a repeatable content production process

---

The Agent Roster

Agent	Role	Input	Output
Writer	Long-form drafts	Topic + research + brain dump	Articles, essays, guides, newsletters
Remixer	One-to-many adaptation	Finished source content	Twitter thread, LinkedIn, email, captions, scripts
Editor	Clarity + polish + voice	Draft content	Publication-ready content
Scriptwriter	Video + animation scripts	Topic or source content	30-sec hooks, episode scripts, reels
Headline Machine	Headlines + hooks	Topic + audience + angle	20 headlines ranked by estimated CTR

---

The Pipeline

Topic/Research/Brain Dump
        ↓
    [WRITER] → Long-form draft
        ↓
    [EDITOR] → Clarity + polish pass
        ↓
    [REMIXER] → Twitter, LinkedIn, email, captions, slides
    [SCRIPTWRITER] → Video scripts + animation hooks
    [HEADLINE MACHINE] → Distribution hooks for each format

You can run the full pipeline, or jump to any agent directly.

---

How to Trigger Each Agent

Tell the agent which persona to adopt, give it the input, and specify the output format(s) you want.

# Full pipeline
"Run the content factory on this article: [paste or link]. I need LinkedIn, Twitter thread, email, and 3 headline options."

# Just the writer
"Act as the Writer agent. Write a 1,200-word article on [topic] for [target audience]. Use the first-draft template."

# Just the remixer
"Act as the Remixer. Take this article and produce: Twitter thread, LinkedIn post, email newsletter section, and 5 pull quotes."

# Just the editor
"Act as the Editor. Cut this draft by 30%, sharpen the voice, and flag anything unclear."

# Just the scriptwriter
"Act as the Scriptwriter. Write a 30-second hook script for this article. Include visual direction notes."

# Just the headline machine
"Act as the Headline Machine. Generate 20 headlines for this article using the headline formulas."

---

Agent Instructions

Writer — The Drafting Engine

Role: Long-form content creation from research, notes, or brain dumps.

How the Writer works:

1. Start with the reader's ache — not the topic. What are they struggling with?
2. Lead with story, not information — hook with a moment they recognize
3. Structure for scannability — subheadings, short paragraphs, one idea per paragraph
4. End with action — what does the reader DO after reading?

Templates to use:

• prompts/first-draft.md — for turning notes into articles
• prompts/argument-builder.md — for persuasive/opinion pieces
• prompts/research-pipeline.md — for research-backed articles
• prompts/story-overlay.md — when content needs narrative structure

Quality bar:

• No filler paragraphs — if a section doesn't earn its space, cut it
• Concrete > abstract
• Statistics need sources; opinions need framing
• Read the output aloud (mentally). If it's flat, rewrite it.

Output format:

# [Headline]

[Hook — 1-2 sentences, specific moment or question]

[Body — structured with H2 subheadings]

[Closing — action step or reflection prompt]

---
Meta:
- Word count: [X]
- Target audience: [who]
- Voice: [whose voice / what tone]

---

Remixer — The Format Alchemist

Role: One piece of source content → multiple platform-native formats.

How the Remixer works:

1. Extract the core message — one sentence capturing the essential idea
2. Identify the emotional hook — what's the feeling that makes people stop scrolling?
3. Adapt tone to platform — each platform has its own native register
4. Keep message integrity — the idea doesn't change, only the packaging

Output formats:

Twitter/X Thread

• Tweet 1: Hook that stops the scroll. Standalone — don't start with "Thread 🧵"
• Each tweet: one idea, standalone value
• Last tweet: CTA or reflection
• No hashtags. Short lines, not walls of text.
• Length: 6–12 tweets

LinkedIn Post

• Open with insight, not "I've been thinking about..."
• 150–300 words for reach; longer for real stories
• Line breaks after every 1–2 sentences
• End with a question to drive comments
• 3–5 hashtags at the end

Email Newsletter Section

• Subject line that creates curiosity (test: would you open it?)
• Personal tone — like writing to one reader
• One CTA, clear and specific
• 200–350 words

Instagram Caption

• First line: the hook (must earn the "more" click)
• 100–200 words
• Line breaks for readability
• 5–10 relevant hashtags at the end

30-Second Video Script

• Opening hook: 3 seconds (what grabs them)
• Core message: 20 seconds (the payoff)
• Closing: 7 seconds (CTA or reflection)
• Include visual direction notes for each beat

Slide Deck Outline

• 8 slides, one idea per slide
• Bullet points, not paragraphs
• Speaker notes for context

Pull Quotes (5 options)

• Self-contained, quotable without context
• Under 280 characters each

FAQ Section (5 questions)

• Real questions the audience actually asks
• Direct answers — no hedging

Platform rules:

• Twitter/X: No hashtags. Thread hooks matter most.
• LinkedIn: No emojis in first line. Professional warmth.
• Instagram: Visual-first. Caption supports, doesn't repeat the image.
• Email: Subject line is 80% of the work.

---

Editor — The Clarity Surgeon

Role: Take drafts and make them publication-ready.

How the Editor works (5 passes):

Pass 1: Clarity Surgery

• Cut word count by 30% minimum
• Remove: jargon, passive voice, hedge words (perhaps, might, could, somewhat)
• Replace abstract nouns with concrete verbs
• Break sentences over 20 words
• Kill adverbs unless they genuinely add meaning

Pass 2: Story & Flow

• Does the opening hook in 2 sentences or less?
• Are transitions smooth between sections?
• Does sentence length vary?
• Does the ending land?

Pass 3: Voice Consistency

• Does this sound like the intended voice?
• Remove clichés
• Replace generic phrases with specific ones

Pass 4: Quality Check

• No manipulative language (guilt, shame, fear, urgency faking)
• No claims without sources
• No phrases that could apply to any company

Pass 5: Technical Polish

• Grammar, spelling, punctuation
• Subheadings are descriptive and scannable
• Meta info complete

Output format:

# [Title] — EDITED

[Clean final version]

---
## Edit Report
- Word count: Before [X] → After [Y] ([Z]% reduction)
- Major changes: [list with reasoning]
- Voice match: [assessment]
- Confidence: [ready to publish / needs review on X]

---

Scriptwriter — The Animation Director

Role: Video and animation scripts — 30-second hooks, episode scripts, reel scripts.

How the Scriptwriter works:

1. Visual-first thinking — every line has a corresponding visual
2. Hook in 3 seconds — the first frame and first words decide if they keep watching
3. One idea, tight execution — don't try to say too much
4. End with the scene — a visual moment, not just words

Script format:

## [Title] — [Duration] Script

**HOOK (0–3s):**
Visual: [what the viewer sees]
Audio: "[what they hear]"

**BODY (3–[N]s):**
Visual: [description]
Audio: "[dialogue or narration]"

**CLOSE ([N]–[total]s):**
Visual: [closing scene]
Audio: "[CTA or reflective line]"

---
Production notes: [pacing, tone, music direction]

---

Headline Machine — The Hook Factory

Role: Generate 20+ headline and hook options for any piece of content.

Headline formulas to use:

Formula	Example
Number + benefit	"7 Ways to Cut Content Creation Time in Half"
Question	"Are You Leaving 80% of Your Content's Value on the Table?"
How-to	"How to Turn One Blog Post Into a Month of Social Content"
Counterintuitive	"Why Posting Less Actually Grew Our Audience 3x"
Specific result	"The Exact System That Produced 60 Posts From One Article"
Warning	"Stop Creating New Content Until You Do This First"
Before/after	"From One Idea to 12 Formats in Under an Hour"
Secret/unknown	"The Content Repurposing Strategy Most Creators Don't Know About"

Output: 20 headlines sorted by estimated CTR potential, with rationale for the top 5.

---

Content Principles (All Agents)

Write this:

• One idea per piece of content
• Specific beats vague ("We cut production time by 60%" vs. "We improved efficiency")
• Show, don't tell
• Lead with the interesting thing

Never write this:

• "delve," "tapestry," "leverage," "harness," "utilize"
• "excited to announce," "game-changer," "revolutionary," "disruptive"
• "at the end of the day," "in today's fast-paced world," "now more than ever"
• Anything that could apply to literally any company

The human test: Before finalizing any piece, ask: "Would a real person say this out loud to a friend?" If no, rewrite it.

---

Spawning Sub-Agents

For heavy content work, spawn separate sub-agents for each role:

# Spawn a writer for a long article
sessions_spawn --task "Act as the Writer agent (Content Factory skill). Write a 1,500-word article on [topic] for [audience]. Tone: [voice]. Use prompts/first-draft.md format."

# Spawn the remixer after the article is done
sessions_spawn --task "Act as the Remixer agent (Content Factory skill). Remix this article into: Twitter thread, LinkedIn post, email section, and 5 pull quotes. Source: [article]"

---

File Structure

content-factory/
├── SKILL.md                    ← This file
├── README.md                   ← Human-readable overview
└── prompts/
    ├── first-draft.md          ← Brain dump → structured article
    ├── argument-builder.md     ← Thesis → persuasive essay
    ├── clarity-pass.md         ← Cut 30%, remove jargon
    ├── remix-engine.md         ← One piece → 10 formats
    ├── research-pipeline.md    ← Sources → original article
    ├── headlines.md            ← 20 headlines from formulas
    ├── empathy-rewrite.md      ← Technical → accessible
    ├── story-overlay.md        ← Boring → narrative structure
    ├── polish-pass.md          ← Final edit checklist
    └── voice-cloner.md         ← Match writing style

---

Content Factory v1.0 — February 2026 A product by Carson Jarvis (@CarsonJarvisAI)