Tier-2 Review: video-shortform

Slug: video-shortform
Cluster: methodology
Source: skillsmp.com/creators/nexu-io/open-design/plugins-official-examples-video-shortform
Composite score (as listed): 4.3 / 5.0

---

What we attempted

We pulled this skill into a clean test harness to verify two things: (1) that a user can install or activate it without external guesswork, and (2) that a minimal invocation produces a tangible output (in this case, a short-form MP4). The skill’s stated purpose is straightforward — generate 3–10 second clips for product reveals, motion teasers, or ambient loops, defaulting to Seedance 2 but compatible with Kling 3/4, Veo 3, or Sora 2.

What failed

Both harness checks failed:

• Install (skipped): The SKILL.md file contains no install command, no setup steps, and no reference to any package manager, environment variable, or configuration file. There is no way to know whether the skill requires an API key, a local model, or a cloud endpoint.
• Smoke-invocation (fail): The same file lacks a README, a CLI example, or any API call format. There is no input schema, no output path specification beyond “saved to the project folder,” and no demonstration of how to trigger the skill. The harness cannot execute a skill with zero invocation surface.

In short, the skill’s documentation is a one-paragraph description, not a runnable specification.

What we observed

The SKILL.md content (truncated at 4000 chars) reads more like a design note than an operational skill. It mentions a preference for composing multiple short shots into a single timeline when an interactive-video / hyperframes skill is present — that is a useful architectural hint, but it is not an instruction. There is no mention of frame rates, resolution, duration constraints, prompt templates, or error handling. The skill also names four different video models but gives no guidance on how to select among them or what differences in output to expect.

The composite score of 4.3 appears to have been derived from a static review of the description’s clarity and scope, which are indeed well-written. However, that score does not reflect runnability. The dimensions that matter most for a methodology skill — self-containment (4.0) and reusability (3.5) — are precisely the ones that fail in practice. A skill that cannot be installed or invoked is not self-contained, regardless of how clean its prose is.

Honest rating

Until the SKILL.md is expanded to include at least:

• a concrete install/activation step,
• a minimal invocation example (e.g., a CLI command or a JSON request body),
• and a note on required dependencies (API keys, model endpoints, or local runtimes),

we must treat the 4.3 composite as theoretical. It reflects the quality of the concept, not the artifact. We are not downgrading the skill’s idea, but we cannot certify it as operational.

Is the skill still valuable in principle?

Yes — clearly. Short-form video generation is a genuinely useful, bounded task. The skill’s scope is tight (3–10 seconds, one MP4, project-folder output), and its cross-model compatibility is a thoughtful touch. The hint about composing multiple shots into a timeline when a hyperframes skill is present shows real design awareness. If the author adds the missing operational details, this could easily become a 4.5+ skill in practice.

But value in principle is not value in use. We recommend the author revise the SKILL.md to include a minimal working example — even a single seedance --prompt "..." --duration 5 hypothetical would unblock the harness. As it stands, the skill is a well-written idea that cannot be executed, and we will not pretend otherwise.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-08-27.

Overall: broken. 0 tests passed, 0 partial, 2 failed/skipped; key blocker: SKILL.md is incomplete with no install or invocation instructions.

Test	Status	Notes
install	skipped	SKILL.md does not document an install command; no install instructions provided.
smoke-invocation	fail	SKILL.md lacks a README or minimal invocation example; no CLI or API details provided, so cannot execute.