Content
56%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.
Strong operational guidance — the preview-vs-run-vs-deploy decision rules, metadata-sync workflow with dry-run and diff feedback loops, and the concrete Java Main-class and //requirements examples are genuinely useful. However, the body mixes large amounts of language-generic Windmill material into a Java-specific skill, is written in long dense paragraphs, and omits the Java-specific parameter/resource details its own General Principles promise.
Suggestions
Move the language-generic guidance (General Principles, Preprocessor Scripts, non-Java parameter syntax) into a shared reference file (e.g. references/general.md) referenced from this skill, keeping SKILL.md focused on Java.
Add the missing Java-specific details the body promises: how a Java script declares resource-typed parameters and how to use the Windmill client (wmill) from Java.
Compress the 'Keep metadata in sync' section into a short numbered workflow (edit → generate-metadata → diff .lock → report version changes) instead of four dense paragraphs.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably verbose for a Java-specific skill: four dense paragraphs on metadata sync, and substantial content irrelevant to Java such as SQL placeholder syntax ("$1 for PostgreSQL, ? for MySQL/Snowflake, @P1 for MSSQL... positional $1, $2, … for Bash, param(...) for PowerShell"), TypeScript/Python resource patterns ("RT.C<Name> in TypeScript, a c_<name> TypedDict in Python"), and preprocessor scripts "written in TypeScript or Python". Not 3: the volume of non-Java material and run-on prose exceeds 'some unnecessary explanation'; not 1: the CLI and Java-specific material is substantive guidance, not padded explanation of concepts Claude already knows. | 2 / 5 |
Actionability | Mostly executable guidance with concrete commands ("wmill script preview <script_path>", "wmill generate-metadata --dry-run", "wmill resource-type list --schema") and a copy-paste-ready Java example plus the //requirements Maven format. Not 5: General Principles promises "the language section below shows how that language declares parameters" and gives TypeScript/Python resource-parameter forms, but the Java section shows only plain String/int parameters — no Java resource-typed parameter or wmill-client usage. | 4 / 5 |
Workflow Clarity | Clear decision rules and checkpoints: preview as the local validation step before deploy, "Only suggest/run a deploy when the user explicitly asks to deploy/publish/push", "--dry-run — it lists each stale item with a reason ... without changing anything", and "diff the regenerated .lock / .script.lock files and tell the user which dependency versions changed" as an explicit feedback loop. Not 5: the write → preview → generate-metadata → deploy sequence is never laid out as an ordered workflow; the reader must assemble it from dense prose. | 4 / 5 |
Progressive Disclosure | There is real structure (## CLI Commands with ### subsections, # Java) and clearly signaled external pointers ("see the **Deploying** section in `AGENTS.wmill.md`", "use the `preview` skill"), but language-generic content (General Principles covering TypeScript/Python/SQL/Bash conventions, and the Preprocessor Scripts section) is inlined in a Java-specific skill — content that belongs in a shared reference file. Not 2: structure and navigation are present and functional; not 4: the misplaced non-Java material is more than a minor organization gap. | 3 / 5 |
Total | 13 / 20 Passed |