Agent Infrastructure  ·  Curated marketplace

brand-extract

Extract a complete Brand Kit from a live website by driving the in-app browser.


Composite

4.3

C 4.3 · A 0.0

How we got there

Craft · D1–D5

D1 · Trigger clarity 4.5
D2 · Output specificity 4.5
D3 · Scope precision 4.5
D4 · Self-containment 4.0
D5 · Reusability 3.5

02 — Review

Our evaluation


Tier-2 Review: brand-extract (nexu-io/open-design)

Slug: brand-extract
Cluster: agent-infrastructure
Source: skillsmp.com listing
Composite score (as listed): 4.3 / 5.0


What we attempted

We tried to validate this skill in a clean container with two standard checks: (1) install — whether the SKILL.md documents a reproducible setup path, and (2) smoke-invocation — whether the skill can be triggered with a minimal example and produce a meaningful output without external manual steps. The harness simulated a fresh environment with no pre-existing browser state or third-party tools beyond the base agent runtime.

What failed

Both checks failed. The install step was skipped because the SKILL.md provides no install command, no dependency list, and no environment setup instructions. The smoke-invocation step failed outright because the skill’s stated trigger assumes a live browser tab and the agent-browser tool — neither of which exists in a clean container. The skill literally says it "pairs with the agent-browser tool" and expects "a site in the Browser tab." Without that tool and tab, there is no code path, no fallback, and no error-handling strategy to degrade gracefully.

What we observed (closer look at the SKILL.md)

The SKILL.md is well-written in terms of intent. It clearly describes when to fire: when a user asks to "extract a brand," "pull the brand from <url>," or "get the colors/fonts/logo from this site." The output spec is also promising — a complete Brand Kit (colors, fonts, logo, likely spacing/tone) measured from the live DOM. The scope is narrow and sensible: it does not try to crawl entire sites or handle non-brand content.

However, the skill is not self-contained. It depends on:

  • An external agent-browser tool (undefined in the SKILL.md)
  • A live browser tab with the target site already open
  • Interactive pauses for anti-bot walls (which is fine, but no timeout or fallback is documented)

There is no README, no minimal invocation example, no "how to install" section, and no note on what happens if the browser tool is missing. This makes the skill untestable in isolation and effectively non-portable.

Honest rating caveat

The 4.3 composite score is theoretical — it reflects the skill’s design quality (clear triggers, specific output, focused scope) but not its operational viability. Until a physical re-run with the agent-browser tool and a real browser tab resolves the failures, we cannot confirm that the skill works end-to-end. The score should be treated as a design review, not a functional certification.

Is the skill still valuable in principle?

Yes — in principle, strongly. Brand extraction from a live site is a real, recurring need in design and front-end work. The trigger phrases are natural, the output (a structured Brand Kit) is exactly what a designer or developer would want, and the scope is tight enough to avoid bloat. The skill’s author clearly understands the domain and the user’s mental model.

The problem is not the idea; it’s the packaging. The skill assumes a proprietary environment (agent-browser + browser tab) without documenting how to obtain or configure that environment. That makes it a good concept prototype but a poor reusable artifact in its current form.

Bottom line: If the author adds a minimal install guide, a fallback for missing browser tooling (e.g., curl + HTML parsing), and a concrete invocation example, this skill could easily justify a 4.5+ rating. As-is, it remains a promising but unverified draft.

03 — Tests

What we tried


Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-08-13.

Overall: broken. 0 tests passed, 0 partial, 2 failed/skipped; key blocker: SKILL.md provides no install or invocation instructions and depends on external browser tool.

Inferred dependencies: agent-browser tool, browser tab.

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; requires agent-browser tool and browser tab, not available in clean container.
04 — Cross-validation

1 source verified

Install

Use this skill

/plugin install brand-extract