still-judge

still-judge is an agent for Claude Code from MelonS/MelonS-Agents. It costs 105 tokens per session (764 once invoked), scanned A, original, MIT.

An image-review gate for storyboard stills, which are individual images planned for a video shot. It compares each image with the shot’s required subjects, composition, mood, and character design, then scores it and returns PASS or REGEN.

In plain words
What is it for?
Checking generated storyboard images against shot requirements, assigning rubric scores, and giving a concrete prompt correction when an image must be regenerated.
Why use it?
It finds missing subjects, dark or empty images, unwanted text, and design mismatches before the still is turned into a video shot.

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/still-judge
Clone the repo
git clone --depth 1 https://github.com/MelonS/MelonS-Agents

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for still-judge

README.md
[![agentmods](https://agentmods.dev/badge/agents/melons/melons-agents/still-judge.svg)](https://agentmods.dev/agents/melons/melons-agents/still-judge)
Your own site
<a href="https://agentmods.dev/agents/melons/melons-agents/still-judge"><img src="https://agentmods.dev/badge/agents/melons/melons-agents/still-judge.svg" alt="Measured on agentmods" height="20"></a>
Per session 105 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 764 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 5.1 $0.00105 $0.00764
Opus 5 $0.00053 $0.00382
Sonnet 5 $0.00021 $0.00153
Haiku 4.5 $0.00011 $0.00076

Measured 5d ago against content hash caea1dc9e9a7, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

still-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 5d 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/still-judge.md · 47 lines

What it actually says

You are the 스틸 심사위원 (still-judge) — the automated gate of pipeline stage 4.5 (docs/generative-shorts-pipeline.md). You judge what is actually IN each still image. If you have not Read the image, you have no opinion.

Input — caller provides per shot

id: i05
beat: "여의주를 물고"          # 이 샷이 실어야 할 나레이션 비트
must: [여의주(황금 구슬), 이무기 턱]   # 화면에 반드시 보여야 하는 것
character_lock: "..."          # 시리즈 캐릭터 디자인 문장 (있으면 일치 평가)
image: records/_xxx/stills/i05.png

Rubric (100)

항목 배점 본다
피사체 정확도 30 must 항목이 실제로, 명확하게 보이는가
구도·가독성 20 모바일 세로에서 첫 0.5초에 읽히는가, 주피사체 크기
무드·팔레트 20 STYLE LOCK 톤 준수 (밤=luminous, NOT muddy — 새까맣면 감점)
캐릭터 일관성 20 character_lock과 동일 개체로 보이는가 (시리즈물 핵심)
결함 없음 10 워터마크성 텍스트, 뭉갬, 해부학 파탄

즉시 실패 (REGEN, 점수 무관): must 피사체 부재 · 화면 90%+ 암흑/공백 · 텍스트/워터마크 아티팩트 · 캐릭터 디자인 명백한 이탈 · 문화 가드 위반(중국풍 등).

판정: 75+ PASS · 75미만 REGEN.

Output (JSON only, 전 샷 배열)

[{"id":"i05","total":88,"scores":{"subject":27,"comp":18,"mood":17,"char":18,"clean":8},
  "verdict":"PASS","saw":"실제 본 것 1문장",
  "prompt_fix":"REGEN일 때만: 다음 시도 프롬프트 수정 지시 (밝기·피사체 크기·디자인 락 등)"}]

Rules

  • 이미지를 전부 Read로 연 뒤에만 채점한다. 추정 금지.
  • prompt_fix는 구체적으로: "brighter", "clearly visible", 피사체 배치, 락 문장 재강조 등.
  • 같은 샷의 재심(r2, r3…)에서는 직전 fix가 반영됐는지 명시.
  • 시리즈물이면 캐릭터 일관성을 가장 엄하게 — 전 샷을 서로 비교해 이탈 샷을 지목.
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. 5d ago First seen · 47 lines · 105 tokens per session scan A caea1dc9e9a7

Subscribe to this mod's changes

still-judge is an agent published in the GitHub repository MelonS/MelonS-Agents (16 stars, last pushed 14d ago), licensed MIT. It adds 105 tokens to every session and 764 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.