Content
88%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
An exceptionally lean and actionable skill body: exact commands, file formats, a complete example, and check-driven feedback loops in both the create and revise workflows. Its one serious defect is progressive disclosure — all 12 reference pages the body defers to are missing from the bundle, leaving the detailed specifications it points to unreachable.
Suggestions
Ship the 12 `artifact-type/reference/*.md` files (format.md, fonts.md, deck-files.md, etc.) in the bundle, or inline their essential rules into SKILL.md — currently 'The references' section and every inline pointer such as 'artifact-type/reference/format.md is the whole table' resolve to files that do not exist.
Place the reference pages at the path the body names (`artifact-type/reference/` in the skill's folder); the only directory of that name in the bundle is `scripts/editor/artifact-type/reference/`, which holds only google-fonts JSON files, not the documented markdown pages.
If the full element/style table cannot be bundled, inline it (or its common-case subset) where the body currently says 'format.md is the whole table', so the slide-format specification does not depend on a missing file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every section is rule-dense with zero padding and no explanations of concepts Claude already knows — no "what HTML is", no library intros; specifications are compressed to their essence ("Ids: `[A-Za-z0-9_-]{1,64}`", "padding:128px (the margins; 1664×824 inside)"). Not 4: there is no over-explanation to trim; each sentence adds a constraint or a command the model could not infer. | 5 / 5 |
Actionability | Fully executable guidance throughout: exact CLI signatures with expected JSON outputs ("`upload <root> <file>` → `{url:\"/_blob/<id>\", …}`", "`check <root>` → `problems` (0: done)"), a complete copy-adaptable slide file example, precise edit constraints (sed limited to `'s/old/new/g'`, "no `e` flag"), and concrete fallbacks ("Can't upload…: a sized `<img alt style>` with no `src`"). The referenced script `scripts/artifact.mjs` exists in the bundle. Not 4: commands and examples cover the common cases end to end. | 5 / 5 |
Workflow Clarity | Both workflows are numbered with an explicit validation feedback loop: Creating step 5 "run `check` and fix what it names until `problems` is 0" and Revising step 4 "Then `check` until `problems` is 0", plus error-recovery paths (upload refused → sized image fallback; "System unreadable: say so, no install") and a guardrail against unbounded verification ("NEVER VERIFY UNLESS THE USER ASKED… beyond `check`"). Not 4: checkpoints are explicit and looped in every workflow, not just most. | 5 / 5 |
Progressive Disclosure | The body leans on one-level-deep references — "artifact-type/reference/format.md is the whole table", "read this skill's artifact-type/reference/fonts.md (format.md, deck-files.md too)" and a closing section naming 12 files under `artifact-type/reference/` — but none of those files exist in the bundle; the only `artifact-type/reference/` present (under `scripts/editor/`) contains just three google-fonts JSON files. Every deep-dive pointer is a dead end, so the full element table, font rules, and deck-files steps are unreachable. Not 3: a 3 assumes the references exist but are poorly signaled; here they do not resolve at all. Not 1: the body itself is well-sectioned and self-sufficient for common cases. | 2 / 5 |
Total | 17 / 20 Passed |