Content
92%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 well-engineered skill body: dense routing table with executable commands, explicit validation gates and acceptance criteria, and a clean overview-to-references split where every referenced file exists. The only notable inefficiency is the near-verbatim duplication of the frontmatter's art-vs-data routing inside the Art source section.
Suggestions
Trim the Art source section's opening ("Image generation is the default for map art ... is data you write and validate, not art") to just the deltas over the frontmatter description — the art_source values, route precedence, and recording rule — since the description is always in context alongside the body.
The Pipeline table enumerates exhaustive flag lists that "--help" already documents per tool; trimming each row to the common-case invocation (e.g. `extract_prop_pack.py --input <sheet> --output-dir <new>` plus one or two high-value flags) would cut tokens while keeping every route actionable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence (no explanations of what Tiled, parallax, or collision are; every section is skill-specific contracts and routing), but the Art source opening ("Image generation is the default for map art: terrain and tile sheets, prop kits, HD-2D plates, painted backgrounds and parallax layers. Map data (layout, collision, exits, spawns) is data you write and validate") re-states the frontmatter description's routing nearly verbatim, and the "only on request" codeart2d routing recurs across four sections. This fits anchor 4 ("Efficient; minor instances ... that could be trimmed") better than anchor 5 ("every token earns its place") because of that duplication, and better than anchor 3 since there is no unnecessary explanation of known concepts. | 4 / 5 |
Actionability | Every route is a copy-paste-ready command with concrete flags and placeholders whose conventions are defined ("extract_prop_pack.py --input <sheet> --rows R --cols C --labels a,b --output-dir <new>", "map_nav.py check --bundle <b> --output-dir <new>"), supplemented by exit-code semantics ("errors print `error: ...` and exit 1; usage errors exit 2") and output guarantees. This meets anchor 5 ("Fully executable; copy-paste ready code or commands; specific examples cover the common cases"); anchor 4 would require missing key details, and none are evident. | 5 / 5 |
Workflow Clarity | The pipeline is explicitly sequenced ("Block out navigation, make art and kit, place it, validate the data, preview, then export"), gated by explicit validation ("A playable map is accepted only after `map_bundle.py validate` and `map_nav.py check` pass"), with feedback loops and error recovery ("regenerate on failure", the stage-beside-and-publish-only-after-checks model, "never say 'all checks passed' when any published envelope has a warn") and an Acceptance checklist. This matches anchor 5 ("Clear sequence with explicit validation steps; feedback loops for error recovery; checklists"); anchor 4's "minor validation gaps" do not apply since every artifact type has a named validator. | 5 / 5 |
Progressive Disclosure | The body is a genuine overview: all twelve referenced .md files, references/schemas/map.schema.json, and references/runtime/map-runtime.mjs exist as real one-level-deep bundle files, each signalled at its point of use ("Choose by behaviour ([map-strategies.md]...)") and again in a consolidated References section. Content that belongs in references (contracts, per-mode guides, schemas) is indeed in them, matching anchor 5 ("Clear overview with well-signaled one-level-deep references ... easy navigation"); there is no nesting beyond one level and no reference wall inlined in SKILL.md. | 5 / 5 |
Total | 19 / 20 Passed |