eval-scenes SKILL.md skillA

eval-scenes from thesampadilla/montaj is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add thesampadilla/montaj/eval-scenes@git:20260930.c29c3f1

Pins git:20260930.c29c3f1 in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
thesampadilla/montaj
Path
skills/eval-scenes/SKILL.md
Versions recorded
4
Latest version
git:20260930.c29c3f1, committed Sep 30, 2026
First version recorded
Apr 21, 2026
Size
9,653 bytes
Repository stars
27
License
MIT

Quality evaluation rubric + retry loop for ai_video generated clips. Load after generating scenes in Phase 6 to evaluate and optionally regenerate clips that fail quality checks.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# Eval Scenes

Evaluate generated scene clips against a 5-dimension quality rubric using Gemini. If a clip fails, regenerate it via `kling_generate` (non-deterministic re-roll with the same composed prompt) and re-evaluate. Repeat up to a retry budget.

This is **Step F** of the ai-video-generate Phase 6 contract — optional but recommended.

## When to load

After all scenes have clips on `tracks[0].items` (Phase 6 Step E complete). The agent runs this loop before setting `status: "draft"`, or after draft if the user requests quality review.

## The rubric

Evaluate each clip on 9 dimensions, scored 1-5. **Pass/fail thresholds vary by dimension** — some are strict (must be >= 4), others are advisory (noted but not gates).

**Gate dimensions (must ALL be >= 4 to pass):**

| Dimension | What to check | Automatic FAIL (score 1-2) |
|-----------|--------------|---------------------------|
| **CHARACTER_MATCH** | Characters match ref images: correct species, colors, proportions, clothing. Specs describe IDENTITY not pose — do not penalize expression/pose differences from the spec. | Wrong species, wrong colors, character morphs into something else |
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add thesampadilla/montaj/eval-scenes@sha256:41abf5d740a26f04

The line at the top of this page pins the label git:20260930.c29c3f1, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_67xtutoqykazazeq.svg)](https://markdownregistry.com/a/art_67xtutoqykazazeq)

4 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 4 versions of this file, between Apr 21, 2026 and Sep 30, 2026, about one every 54 days.

versioncommittedcommitsizeaudit
git:20260930.c29c3f1 latest2026-09-30 c29c3f1 9,653 BA view · diff
git:20260822.ed4dd4e2026-08-22 ed4dd4e 9,645 BA view · diff
git:20260423.fad7e8e2026-04-23 fad7e8e 9,621 BA view · diff
git:20260421.84b28fc2026-04-21 84b28fc 9,612 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (9653 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

thesampadilla/montaj · 27 stars · license MIT · pushed 2026-10-19 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_67xtutoqykazazeq
GET https://markdownregistry.com/api/v1/resolve?ref=thesampadilla/montaj/eval-scenes
GET https://markdownregistry.com/api/v1/blob/41abf5d740a26f0427144e8718a46e722848ab3f296fc3cc3ffb0877b3f89ec5

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from thesampadilla/montaj

montaj SKILL.md skill
thesampadilla/montaj · skills/SKILL.md · You MUST use this whenever the user asks for video editing work. Use it when video-related tasks are brought up…
git:20261013.016a16d · audit A · 27 stars
_contract SKILL.md skill
thesampadilla/montaj · skills/_contract/SKILL.md · The shared vocabulary that domain skills use to name Montaj operations. Domain skills phrase every Montaj interaction…
git:20260920.87073e5 · audit A · 27 stars
ai-video-generate SKILL.md skill
thesampadilla/montaj · skills/ai-video-generate/SKILL.md · Generation orchestration for approved ai_video storyboards: dispatch mode selection, scene generation via…
git:20261015.b7d88f3 · audit A · 27 stars
ai-video-plan SKILL.md skill
thesampadilla/montaj · skills/ai-video-plan/SKILL.md · Agent-authored director skill for the ai_video workflow: story clarification, storyboard writes, approval gate. Load…
git:20260930.c29c3f1 · audit A · 27 stars
camera-vocabulary SKILL.md skill
thesampadilla/montaj · skills/camera-vocabulary/SKILL.md · Shot scale and camera move values, and the motion-budget rule, for ai_video scene planning. Load when writing…
git:20260927.c9cc65a · audit A · 27 stars
carousel SKILL.md skill
thesampadilla/montaj · skills/carousel/SKILL.md · Agent-authored workflow task: build slides for a carousel project — pick aspect-aware images via generate_image…
git:20261007.0741b8f · audit A · 27 stars
edit-session SKILL.md skill
thesampadilla/montaj · skills/edit-session/SKILL.md · Loaded after a draft is ready for back-and-forth editing: cutting, overlay adjustments, timing fixes. Not a pipeline…
git:20261004.27b544d · audit A · 27 stars
editable-text SKILL.md skill
thesampadilla/montaj · skills/editable-text/SKILL.md · Author carousel text overlays whose font/size/color/etc. the operator can edit from the editor toolbar. Load when…
git:20261008.376147f · audit A · 27 stars
image-search SKILL.md skill
thesampadilla/montaj · skills/image-search/SKILL.md · Find and download real images from the web — people, logos, brand/event stills, B-roll — to add as image clips on the…
git:20261007.0741b8f · audit A · 27 stars
mcp SKILL.md skill
thesampadilla/montaj · skills/mcp/SKILL.md · MCP server mode for montaj — load when running as an MCP client (e.g. Claude Desktop)
git:20260930.5b34692 · audit A · 27 stars
native SKILL.md skill
thesampadilla/montaj · skills/native/SKILL.md · The native interface — how the _contract verbs are performed when Montaj runs locally, either via the HTTP server…
git:20260920.87073e5 · audit B · 27 stars
onboarding SKILL.md skill
thesampadilla/montaj · skills/onboarding/SKILL.md · A guided first session in the Montaj app: what Montaj is, checking the connection, then walking the user through their…
git:20261009.9add511 · audit A · 27 stars

Every file in thesampadilla/montaj

Browse by kind, by grade A, or by owner.