skills.rest
dimagi-internal
dimagi-internalOFFICIAL
Explore all skills by this author →
GitHub

partnership-video-build-eval

Grades video build artifacts on four weighted quality dimensions and outputs verdict YAML.

1|2|Updated Apr 1, 2026
Search Tags:#llm-as-judge#brand safety#quality grading#content grounding#render validation#partnership video evaluation#prospect-facing content
One-click install
npx skills add https://github.com/dimagi-internal/ace --skill partnership-video-build-eval
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: partnership-video-build-eval
Source: https://github.com/dimagi-internal/ace/tree/main/skills/partnership-video-build-eval
Command: npx skills add https://github.com/dimagi-internal/ace --skill partnership-video-build-eval

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Manually reviewing partnership video build artifacts for quality, factual accuracy, brand safety, and render success is time-consuming and prone to oversight, risking prospect-facing content with uncited claims, unauthorized branding, or broken renders.

Core Features & Use Cases

  • Four-Dimension Quality Grading: Evaluates video specs across weighted dimensions: spec validity (20%), factual grounding (45%), render success (20%), and brand safety (15%).
  • Hard Safety Guardrails: Automatically flags and blocks artifacts with uncited narration claims, unauthorized third-party branding, or confirmed render failures to prevent risky content from reaching prospects.
  • Standardized Verdict Output: Writes a structured verdict YAML that aggregates with other evaluation results, and is gated by upstream partnership-video-build inline QA to avoid grading incomplete artifacts.
  • Use Case: ACE operators building prospect partnership videos can use this skill to automatically validate that all content is traceable to research, compliant with brand rules, and rendered correctly before proceeding to publication.

Quick Start

Use the partnership-video-build-eval skill to grade the latest partnership video build artifact for your target prospect and generate the required verdict YAML for the run folder.

Frequently Asked Questions about partnership-video-build-eval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate partnership video builds for factual grounding and brand safety?▼

Uncited narration claims, unauthorized third-party branding, and confirmed render failures act as hard blockers. The evaluation enforces safety guardrails by automatically flagging these issues to prevent risky content from reaching prospects.

How does LLM-as-judge content grading work for prospect-facing video validation?▼

Quality dimensions are weighted as factual grounding at 45%, spec validity at 20%, render success at 20%, and brand safety at 15%. This weighting prioritizes traceable research and content accuracy above all other factors.

How do I validate video specs and check for uncited claims before publication?▼

Yes, the evaluation is gated by upstream partnership-video-build inline QA to avoid grading incomplete artifacts. This ensures only finished builds proceed to the four-dimension quality grading and verdict YAML output.

What are the limitations of automated render validation for partnership videos?▼

The evaluation outputs a standardized verdict YAML file written to the run folder. This structured format allows the results to aggregate seamlessly with other evaluation data in the opp-eval pipeline for downstream analysis.

Related Skills

Explore More Marketing & Sales Skills→
dimagi-internaldimagi-internal
1

partnership-microdemo-eval

Grade partnership micro-demo clips for fidelity, relevance, and provenance honesty.

Official
Advanced
jjacksonjjackson
1

ocs-chatbot-eval

Processes chestnut images and generates a PDF report of findings.

Community
Advanced
dimagi-internaldimagi-internal
1

video-spec-generate

Generate validated spec.yaml files for ace-web video programs from source content.

Official
Advanced
gaos6egaos6e

qa-gate

Validate artifacts for factual accuracy, tone, completeness, and data sensitivity.

Community
Advanced