Introduction

Imagine telling a friend "write me something funny about space communists" and receiving a polished, four-minute comedy sketch complete with stage directions, casting notes, and a running gag about procedural spacing protocols. That is what happened when we fed two-sentence prompts into an AI fan fiction generator built on Augment Extensions. The results were genuinely funny — not the cringe AI humor you expect.

The secret is modular context injection. Augment Extensions breaks through the ~49,400 character limit that constrains AI coding assistants like Augment Code AI. The screenplay writing module injects comedy structure patterns, character voice templates, and SNL-specific format rules directly into the AI's context window. This transforms a general-purpose language model into a specialized comedy writing agent that understands setup, escalation, and callback timing.

From a computational creativity perspective, this demonstrates emergent creative behavior through structured prompt engineering. Domain-specific context modules transform large language models from generic text generators into specialized creative agents. The implications extend beyond comedy writing — any domain with complex rules and conventions benefits from this modular approach to AI-first application development.

The Experiment — Testing AI Beyond Code

I built Augment Extensions to solve a real developer problem: AI coding assistants that hallucinate because they lack sufficient context. The npm package ships with 20+ modules covering Python, TypeScript, React, and more. But a persistent question nagged me — does modular context injection work for creative projects too?

SNL sketches make the perfect test case. They are short (3-4 minutes), self-contained, and humor-dependent. Comedy is the hardest creative writing challenge for AI. Generic chatbots produce bland, predictable jokes that read like a dad joke generator crossed with a corporate memo. If Augment Extensions could produce genuinely funny sketches, the system works for anything.

The hypothesis was simple. If feeding an AI structured coding standards produces better code, then feeding it structured comedy writing standards should produce better comedy. I wanted something easy, short, contained, and fun to read. Classical sketch comedy structure — setup, escalation, callback — maps cleanly onto a module's rule format. So I built the screenplay modules and tested them with three absurd premises.

How the Screenplay Modules Enable AI Fan Fiction

The Augment Code AI platform limits the .augment/ folder to approximately 49,400 characters. That is enough for basic coding standards but insufficient for deep domain expertise. Augment Extensions solves this by loading external modules on demand through the augx CLI tool.

The screenplay writing module contains several key components. Character voice templates define distinct speech patterns, catchphrases, and emotional ranges for each role. Comedy structure patterns enforce the classical sketch format: establish the premise, introduce the first unusual thing, heighten through escalation, then resolve with a callback or blackout. Running gag generators create recurring jokes that build throughout the sketch.

Here is the module configuration that powers the screenplay generation:

```json
{
  "name": "screenplay-writing",
  "version": "1.0.0",
  "modules": [
    "comedy-structure",
    "character-voice",
    "snl-format",
    "scene-transitions",
    "running-gags"
  ]
}
```

Linking the module takes one command. The augx link screenplay-writing command injects all comedy writing expertise into the AI's context. The AI then understands SNL-specific conventions: cold opens, host monologue format, Weekend Update style, and recurring character patterns. This domain knowledge transforms generic AI output into production-ready creative content.

The Three Sketches — From Prompts to Complete Scripts

Each sketch started with a two-sentence prompt. No character descriptions, no plot outlines, no dialogue suggestions. Just a premise and a tone. The AI screenplay generator handled everything else — casting, stage directions, running gags, and comedic timing. You can read all three complete scripts on the screenplay reader page.

🚀 SPACE COMMIES — "Spacing Is Just Good Policy"

The prompt was deliberately absurd: a forgettable Starfleet vessel where the Captain decides to space some communist prisoners. The AI generated a complete 3:30 cold open with five characters — The Captain, First Officer, Engineer, Ensign, and Admiral. The First Officer objects on procedural grounds, but the Captain's logic is so airtight that he eventually agrees spacing commies is just good policy.

What impressed me most was the escalation. The sketch builds through increasingly ridiculous bureaucratic justifications. The running gag of treating prisoner spacing as standard paperwork carries the entire piece. The Admiral's video call entrance in Act III adds a perfect authority figure twist. Read the full interactive script — the HTML version includes production notes and stage directions.

🔥 SELF-DESTRUCT OVEN — "A Feature, Not a Bug"

This one started with a single absurd premise: FlameTek Industries' self-cleaning oven has caused 847 house fires, affecting 40% of all units sold. The AI built a six-character corporate boardroom sketch with Ron Pemberton III (CEO who does not know what product they make), Brenda Hartford (VP begging for a recall), Derek Chen (CMO calling fires "a feature"), and three more executives enabling the chaos.

The sketch runs 4:00 and follows a product demo format that escalates from denial to delusion. Legal says everything is "alleged." The CMO pitches a marketing campaign around the fires. The CEO reveals he thought they made refrigerators. Every character has a distinct voice and comedic function. The classical comedy sketch structure — premise, complication, escalation, blackout — executes with precision across every scene.

🧙 WIZARD IN THE ENGINE ROOM — "It's Always Been Magic"

The wildest premise: Starfleet's dirty secret is that faster-than-light travel is scientifically impossible. Every ship runs on magic. Wizard Clif High has kept the USS Enterprise flying for 47 years, but when Starfleet threatens his relationship with a young engineer named Heidi, he quits on the spot.

This 3:30 sketch blends workplace comedy with fantasy elements in a way that feels fresh. The Captain, First Officer, and Ensign discover their entire careers depend on a wizard they have been hiding from Starfleet Command. The humor comes from military officers trying to apply regulations to literal magic. Clif High is the standout character — a cranky wizard who treats warp speed like plumbing. The emotional arc between Clif and Heidi gives the sketch unexpected heart.

The Vibe Coding Workflow for AI Fan Fiction

Vibe coding means describing what you want and letting AI handle the implementation. The same approach works for creative writing. Here is the step-by-step workflow for generating your own AI fan fiction with Augment Extensions.

Step 1: Install Augment Code AI in VS Code. The extension provides the AI backbone that processes your prompts and generates content.

Step 2: Clone the augment-extensions repository and install the CLI tool. Run npm install -g @mytechtoday/augment-extensions to get the augx command.

Step 3: Link the screenplay module. Run augx link screenplay-writing in your project directory. This loads comedy structure rules, character voice templates, and format conventions into your AI context.

Step 4: Write your prompt. Keep it simple — two sentences describing the concept and tone. The module handles the rest. Here is an example prompt you can try right now:

```markdown
Write an SNL cold open sketch about a pizza delivery driver
who accidentally delivers to the Oval Office during a cabinet
meeting. The President insists on splitting the check evenly
even though the Secretary of Defense ordered five pizzas.
```

Step 5: Iterate. The first draft is usually 80% there. Use follow-up prompts to adjust pacing, sharpen specific jokes, or add characters. The module maintains consistency across iterations because the comedy structure rules persist in context.

The entire process takes about ten minutes from prompt to polished script. Compare that to the hours or days a human comedy writer spends on a single sketch draft. The AI does not replace the creative vision — it accelerates the execution.

Beyond SNL — Other Fan Fiction Possibilities

The screenplay modules demonstrate a broader principle. Any creative writing format with defined rules and conventions can benefit from modular context injection. The OpenSpec specification framework that powers Augment Extensions supports any domain module you want to build.

Movie scripts follow three-act structure with specific beat timing. Sitcom episodes use A-plot/B-plot architecture with act breaks. Stage plays require different dialogue conventions than screen media. Short stories follow genre-specific narrative patterns. Each of these formats has documented rules that translate directly into Augment Extensions modules.

Building a custom writing module is straightforward. Define your genre's structural rules, create character archetype templates, specify dialogue conventions, and document pacing expectations. The Beads task tracking system helps manage module development across iterations. Once linked, your custom module gives the AI deep expertise in your chosen creative domain.

The implication goes beyond entertainment. Technical writing, marketing copy, legal documents, and educational content all follow domain-specific conventions. Augment Extensions makes the AI versatile across any structured output format. The Model Context Protocol shows how the industry is moving toward this modular, context-aware approach to AI tooling.

What This Proves About AI Creative Tools

The screenplay experiment passed the creativity test convincingly. Three two-sentence prompts produced three complete, distinct comedy sketches with unique characters, escalating humor, and proper dramatic structure. The AI did not just fill in a template — it made creative choices about character dynamics, comedic timing, and thematic callbacks that felt intentional.

The key insight is domain-specific context. Generic AI prompting produces generic results. When you ask ChatGPT or a basic GitHub Copilot setup to write a comedy sketch, you get predictable setups with obvious punchlines. The screenplay module gives the AI understanding of why certain comedy structures work — escalation creates tension, callbacks reward attention, distinct character voices prevent dialogue from blending together.

Here is an honest comparison of the three approaches to AI-assisted creative writing:

Human writers still excel at originality, cultural references, and the kind of observational humor that comes from lived experience. AI handles structure, consistency, and rapid iteration brilliantly. The sweet spot is collaboration — use AI to generate drafts quickly, then apply human judgment to polish the humor and add personal voice. The vibe coding methodology applies to creative writing as much as software development.

What AI Does Well

• Maintains consistent character voices across an entire script
• Applies comedy structure rules (setup, escalation, callback) reliably
• Generates complete production notes, stage directions, and casting suggestions
• Produces first drafts in minutes instead of hours
• Iterates on feedback without ego or writer's block

Where Human Writers Still Excel

• Original observations drawn from personal experience
• Cultural timing — knowing what is funny right now
• Subversive humor that breaks conventions intentionally
• Emotional nuance and vulnerability in comedy
• The instinct to cut a joke that technically works but feels wrong