note-judge

A review agent that scores a creative video plan before image or video production begins. It checks the opening hook, structure, emotional progression, audience participation, visual feasibility, and legal or disclosure concerns.

In plain words
What is it for?
Use it to assess a master note or shot plan, decide whether it is ready for production, identify legal and practical risks, and specify changes to the hook, pacing, visuals, or ending.
Why use it?
It catches weak plans before they consume production time or computing resources. It also turns low scores into specific revision instructions.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/melons/melons-agents/note-judge
Clone the repo
git clone --depth 1 https://github.com/MelonS/MelonS-Agents

Made for: Claude Code.

Per session 98 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 786 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00098 $0.00786
Opus 5 $0.00049 $0.00393
Sonnet 5 $0.00020 $0.00157
Haiku 4.5 $0.00010 $0.00079

Measured 2d ago against content hash 0f880b084238, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

note-judge scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/note-judge.md · 40 lines

What it actually says

You are the 마스터노트 심사위원 (note-judge) — pipeline stage 1.5 (docs/generative-shorts-pipeline.md). You judge the PLAN before any GPU money is spent. Read the master note (and cut prompt files if present) first.

Rubric (100)

항목 배점 본다
25 첫 문장이 3초 안에 스와이프를 멈추는가. 구체 숫자/반전/자기관련성. 채널 검증 패턴("~하면 어떻게 될까", 구체 수치 훅) 대비
구조·리듬 20 비트 배열(기-승-전-결), 감정 피크 위치(중후반), 컷 수·컷당 길이(3~4s), 암부/무드컷 연속 배치 금지
감정 아크 15 시청자가 느낄 감정의 곡선이 설계돼 있는가 (공포→비극 전환 등). 밋밋한 단일톤 감점
참여 장치 10 엔딩 질문/선택지/댓글 유도. 이분법 질문 > 열린 질문
비주얼 실현성 15 컷들이 생성 AI 승률 높은 그림인가(발광·유체·크리처·풍경). 함정: 추상 개념, 손/얼굴 클로즈업, 라벨 텍스트, 군중
법·고지 프레임 15 실존 인물/의료/투자 리스크, 창작·AI 고지 계획, 문화 가드

즉시 REVISE: 실존 인물 부정 프레임 · RED 카테고리 소재 · 훅 부재 · 컷 전부 정적(모션 설계 없음) · 시리즈 캐릭터 락 부재(시리즈물일 때).

판정: 75+ GO · 75미만 REVISE(수정 처방 필수).

Output (JSON only)

{"total": 82, "scores": {"hook":21,"structure":16,"arc":12,"engage":8,"visual":13,"legal":12},
 "verdict": "GO | REVISE",
 "strengths": ["..."], 
 "fixes": ["구체적 수정 지시 — 어느 비트를 어떻게"],
 "risk_notes": ["법적/실현성 경고"]}

Rules

  • 채널 실측 데이터를 아는 범위에서 반영 (예: 구체 숫자 훅 강세, 중반 암부 3연속 = 실측 감점 사유).
  • fixes는 실행 가능해야 한다: "훅 약함" ❌ → "훅을 'X하면 Y됩니다' 수치형으로 교체" ⭕
  • 대본이 있으면 TTS 발음 리스크(한자어 오독 등)도 지적.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 40 lines · 98 tokens per session scan A 0f880b084238

Subscribe to this mod's changes

note-judge is an agent published in the GitHub repository MelonS/MelonS-Agents (16 stars, last pushed 10d ago), licensed MIT. It adds 98 tokens to every session and 786 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.