Takes an existing screenplay or pilot script and builds it into a complete, pitch-ready package.
Five stages: coverage diagnosis → structure audit → logline → pitch documents → verbal pitch. Each stage must pass before the next opens. Fix before you package. Package before you pitch.
Most scripts never get pitched because they're never finished. Not unwritten — unpackaged. The writer has a draft. Maybe several. But no logline that works, no one-pager that earns the meeting, no understanding of what a development executive actually needs to see.
This skill closes that gap. It starts with the script you have and produces the exact materials the industry expects.
| Stage | What it produces |
|---|---|
| 1. Coverage diagnosis | Six-criterion rubric — PASS/CONSIDER/RECOMMEND per criterion. Priority-ranked fix list. |
| 2. Structure audit | Page-by-page check against benchmarks for the script's format. Every broken benchmark identified. |
| 3. Logline | One or two sentences that make an executive want to read more. Formula, tests, ten failure modes. |
| 4. Pitch documents | One-pager, treatment, pitch deck, series bible — each defined exactly and built to spec. |
| 5. Verbal pitch | 2–3 minute structure. Six failure modes. Ready for the room. |
skills/script-finisher/
SKILL.md the five stages and the discipline
references/
coverage-diagnostic.md six-criterion coverage rubric
structure-audit.md act structure benchmarks by format
logline-workshop.md formula, tests, ten failure modes
pitch-documents.md every document in the package, exactly
verbal-pitch.md the 2–3 minute pitch structure
sources.md sources graded [INDUSTRY] / [PRACTITIONER]
git clone https://github.com/Nagacash/Pitch-ready.git
mkdir -p ~/.claude/skills
cp -r Pitch-ready/skills/script-finisher ~/.claude/skills/Paste SKILL.md into the system prompt. Load individual reference files when you need depth on a specific stage.
from pathlib import Path
from openai import OpenAI
skill = Path("skills/script-finisher/SKILL.md").read_text()
client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key="...")
resp = client.chat.completions.create(
model="nousresearch/hermes-4-70b",
messages=[
{"role": "system", "content": skill},
{"role": "user", "content": "I have a rough feature script about a forensic accountant. Help me build the pitch package."},
],
)It does not write the script. It diagnoses, structures and packages one that already exists.
It does not guarantee a meeting or a sale. The concept and execution have to be there — which is why coverage diagnosis happens first.
It does not substitute for a manager or entertainment attorney. Note for AI-assisted work: WGA guidance requires disclosure when AI-generated material is used in a submission context.
Hermes 4 ran the skill against an actual short-form comedy series (30-episode vertical serial) — not a test scenario but a real project in development.
"It forced a format decision before packaging. That's the move a real development exec makes and most 'pitch' prompts skip."
"It caught the one gap a founder can't see from inside: NAGA is short-form-native, not a pilot PDF. A vanity pass would've called it done. This didn't."
"The reference files are real, specific, and industry-grounded (Netflix deck guidance, WGA disclosure note, the escalation override test). Not filler."
"The skill grades structure and format, not funny. It confirmed NAGA is well-built and packageable. It did not and cannot tell you the jokes land — that's a different tool. Honest engineering, not hype."
The test also found the one real gap: structure-audit tables are page-count based (feature, pilot, half-hour). The skill didn't have benchmarks for short-form vertical serials — 2–8 minute episodes in a 10–30 episode run. That format is now covered in references/structure-audit.md.
Result: the draft returned RECOMMEND with one actionable fix — collapse to a formatted pilot PDF. One script away from submission-ready.
- narrative-film-direction — directing methodology for once the script is greenlit
- character-continuity-skill — keeping characters consistent in production
- prove-it — verification discipline for AI agents
Documentation — CC BY 4.0. Attribution: Naga Codex (Maurice Holda) — https://github.com/Nagacash/Pitch-ready
Created by Naga Codex — Maurice Holda, Hamburg.