Tier-2 Review: od-contribute

What we attempted

We set out to run the standard two-stage harness against od-contribute: an install step to resolve any declared dependencies, followed by a smoke invocation to confirm the skill could be triggered and would produce a coherent output. The skill advertises a "one-click contribution flow" for the OpenDesign repository — four contribution cards (ship a Skill or Design System, translate docs, fix a typo / write a blog, report a bug), each backed by an agent that validates input and opens a PR or issue.

What failed

Both harness stages were skipped, and skipped for a specific, diagnosable reason: the supplied SKILL.md text (truncated to 4000 chars for this review, 6000 for the harness) contains no install command, no package name, no dependency declarations, and no README or minimal invocation example. The document opens directly into marketing-style trigger language and a four-card description. There is nothing executable to exercise.

This is not a soft failure. The harness reported 0 passed, 0 partial, 0 failed — every test was skipped before it could run. A skip is not a pass; it is the absence of evidence. We could not confirm the skill installs, that its triggers fire under realistic phrasing, that the agent actually opens a PR, or that its validation behaves as described.

The most likely explanation is structural rather than a defect in the skill itself: the source page (skillsmp.com) appears to expose a truncated front-matter-plus-description view, not the full skill body. If the real SKILL.md elsewhere contains install steps, dependency lists, and invocation examples, our harness never saw them. The failure mode, stated precisely, is an under-specified or truncated artifact that cannot be independently executed from what was provided.

What we observed

The visible content is high-signal on intent and low-signal on execution. Trigger words are explicit and unusually well-chosen ("contribute to open design," "ship my OD skill," "report an OD bug," and the literal slug od-contribute), which is why D1 scored 5.0. Scope is tightly bounded to one repository (nexu-io/open-design), which reads as deliberate rather than vague — hence D3 at 4.5. But output specificity (4.0) and self-containment (4.0) are capped by the same missing material that blocked the harness: we can describe what the skill claims to do, not how it does it or what a caller must supply.

The low reusability score (2.5) is the honest tell. A skill that only makes sense against one upstream repo, and whose invocation contract is not documented in the artifact we received, is hard to lift into another context without re-reading the source.

On the rating

The composite of 4.1 / 5.0 is therefore theoretical. It reflects the quality of the description we could read, not demonstrated behavior. Until the failures above are resolved — specifically, until a full SKILL.md with a documented invocation and any dependencies is available — no reviewer should treat that number as validated. If the truncation is an artifact of the listing page and the real skill is complete, a physical re-run would likely move the rating little on D1/D3 and materially on D4/D5. If the skill genuinely ships without an invocation contract, the rating is optimistic.

Does it still seem valuable in principle?

Yes. The problem it targets is real and underserved: lowering the contribution barrier for non-coders into a design-systems repo is exactly the kind of agent-infrastructure work worth having, and the four-card framing is a clean UX for it. The concept is sound; the packaging is what failed our harness, and that is fixable with a documented install and a single worked invocation example. We would re-test eagerly once that exists — but we will not claim it passed when it did not.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-10-02.

Overall: broken. 0 tests passed, 0 partial, 0 failed; both tests were skipped because the truncated SKILL.md exposes no install command, README, or dependency declarations to exercise.

Test	Status	Notes
install	skipped	The provided SKILL.md content (truncated to 6000 chars) contains no documented install command, package name, or dependency list, so no install step could be executed or verified.
smoke-invocation	skipped	No README or minimal invocation example is present in the supplied SKILL.md text; only trigger words and a four-card description are given, so a smoke invocation cannot be constructed.