Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/WhiteTowerAI/cut-as-codenpx agentmods add skills/whitetowerai/cut-as-code/video-color-gradeWrote 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.
[](https://agentmods.dev/skills/whitetowerai/cut-as-code/video-color-grade)<a href="https://agentmods.dev/skills/whitetowerai/cut-as-code/video-color-grade"><img src="https://agentmods.dev/badge/skills/whitetowerai/cut-as-code/video-color-grade/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/whitetowerai/cut-as-code/video-color-grade"><img src="https://agentmods.dev/badge/skills/whitetowerai/cut-as-code/video-color-grade.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00047 | $0.03015 |
| Opus 5 | $0.00023 | $0.01507 |
| Sonnet 5 | $0.00009 | $0.00603 |
| Haiku 4.5 | $0.00005 | $0.00301 |
Grade A, and why
video-color-grade 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 10d 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.
How it starts
The opening of the file, as written. The whole thing — 219 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Video Color Grade (assess → correct → looks → choose → LUT + apply)
Write the grade from scratch as code, generate several distinct named looks on the real
footage, select one through explicit human choice or delegated agent judgment, then bake a
portable .cube LUT and apply it to the full clip. No preset packs.
Core idea — correct, then style. Every look is corrective base + creative layer:
- corrective base makes the footage honest — for flat LOG/HLG it's the log→Rec.709 conversion; for already-Rec.709 footage it's WB neutralize + exposure + mild contrast. Every look shares this base, so "neutral" stays meaningful and looks differ by intent.
- creative layer is the look stacked on top (warm / cool / teal-orange / punchy / faded…).
Skin is the quality bar. On people — especially darker skin — teal_orange and
cool_desat routinely throw a green/grey cast. Judge every look on faces first; keep midtone
warmth, push the tint into shadows/highlights/background, prefer vibrance over raw
saturation. Don't push the white balance all the way to neutral or skin goes lifeless.
Requirements: ffmpeg/ffprobe on PATH, Python with numpy + Pillow.
Contract
- Every look =
base + creative layer; the base is shared by all looks. - All looks render on the same representative frame (a fair, controlled comparison), labeled, at full resolution.
- The full apply uses
-c:a copy(audio never re-encoded → A/V sync identical) and keeps source duration, fps, dimensions. - Complete look selection before delivery. Use
humanmode when the user asked to choose; useagentmode without pausing when the user delegated the choice or requested an autonomous run. In both modes, preserve the rationale in the durable plan.
Project protocol workflow
Store the durable decision in work/color-grade/grade-plan.json:
{
"schema_version": 1,
"target": "base-video",
"base": "eq=contrast=1.05",
"looks": [{"name": "clean", "chain": "null"}],
"selected_look": "clean",
"selection_mode": "agent",
"selection_rationale": "Neutral correction preserves natural skin and source lighting.",
"selected_lut": "../../final/selected-color-look.cube",
"evidence_refs": ["media:source"]
}
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 10d ago First seen · 219 lines · 47 tokens per session scan A 0fc6d9a5cb71
video-color-grade is a skill published in the GitHub repository WhiteTowerAI/cut-as-code (11 stars, last pushed 17d ago), licensed MIT. It adds 47 tokens to every session and 3,015 once invoked, about $0.0002 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.
Other skills, from other repositories
smart-video-editor
A video-editing workflow that examines raw clips, selects usable moments, arranges them into a story, and renders a finished video with FFmpeg.
create_profile_style_skill
A workflow for turning a person's video-editing habits into a reusable editing skill file. It examines choices such as pacing, storytelling, audio, subtitles, colour, transitions, and recurring editing rules.
subtitle_imitation_skill
A video-writing workflow that rewrites a script to match the style of reference text supplied by the user. It uses the footage analysis and reference sample while keeping the script tied to what is actually shown.
premiere-pro-mcp
Install, verify, troubleshoot, and operate the Adobe Premiere Pro MCP server. Use when a user wants an agent to set up Premiere MCP, connect Claude Code/Codex/Claude Desktop, control Premiere, import media, build sequences, edit timelines, apply effects, or diagnose bridge issues.
ai_transition_editing_skill
A video-editing workflow that connects existing clips with AI-generated transitions. It can organize clips, choose background music, build a timeline, and render the finished video.
default_editing_workflow_skill
A general video-editing workflow for turning uploaded or found footage into a finished video. It covers loading clips, organizing a story, adding text, voiceover, music, and effects, then rendering the result.