video-spec-eval

Grade ace-web video program spec.yaml files against a 6-dimension rubric.

1|2|Updated Apr 1, 2026
One-click install
npx skills add https://github.com/dimagi-internal/ace --skill video-spec-eval
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: video-spec-eval
Source: https://github.com/dimagi-internal/ace/tree/main/skills/video-spec-eval
Command: npx skills add https://github.com/dimagi-internal/ace --skill video-spec-eval

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Manually grading video program specs for quality is subjective and inconsistent, leading to generic, low-impact Connect video content that fails to resonate with viewers and doesn't accurately reflect the source program.

Core Features & Use Cases

  • 6-Dimension Quality Grading: Evaluates video program specs across narration voice, stat selection, beat coherence, source fidelity, tagline mirror, and story compression, with 0-10 scores, concrete strengths/weaknesses, and one-line improvement recommendations per dimension.
  • Prompt-Independent Evaluation: Derives all quality anchors from the template bundle (intent + example spec) and source page, with no reliance on external prompt files for consistent, unbiased grading.
  • Structured Verdict Output: Emits a standardized YAML verdict with an overall pass/revise/fail status, dimension scores, and a summary of the highest-impact edits to make specs production-ready.
  • Use Case: After generating a video spec with /ace:video-spec-generate or /ace:video-from-program-page, run this eval to get an objective quality assessment and specific edits to improve the spec before rendering the final video.

Quick Start

Use the video-spec-eval skill to grade the spec for your Connect video program and receive a structured quality verdict with actionable improvement recommendations.

Frequently Asked Questions about video-spec-eval

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

FAQPage Schema
How do I grade video program spec quality objectively?

A standardized 6-dimension quality rubric grades video program specs by evaluating narration voice, stat selection, beat coherence, source fidelity, tagline mirror, and story compression, assigning 0-10 scores with concrete strengths, weaknesses, and improvement recommendations per dimension.

What's the best way to evaluate a Connect video spec before rendering?

Run a post-generation spec evaluation that applies a 6-dimension rubric to your spec.yaml file, deriving quality anchors from the template bundle and source page to emit a structured YAML verdict with pass, revise, or fail status and specific edits to improve the spec.

Can I use an LLM-as-judge rubric on hand-authored video specs?

Yes, the evaluation rubric applies to hand-authored video specs, as well as those produced by automated generation commands, by applying the same 6-dimension grading criteria and word budget rules to derive quality scores and actionable fixes.

Does video spec evaluation require external prompt files for grading?

No, video spec evaluation is prompt-independent, deriving all quality anchors from the template bundle and source page, ensuring consistent and unbiased grading without relying on external prompt files.

Why does my video spec evaluation return a revise status?

A spec evaluation returns a revise status when the spec.yaml file partially meets the 6-dimension rubric criteria, requiring specific high-impact edits to dimensions like narration voice, beat coherence, or source fidelity before the spec is production-ready for rendering.