Content
75%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.
The body is a strong, highly actionable reference: exact wmill commands, decision rules keyed to user intent, and complete Go examples make it copy-paste executable. Its weaknesses are repetition (deploy-gating and 'not a deploy' stated multiple times) and multi-language scope creep that pads a nominally Go-only skill, plus no guidance on handling preview failures.
Suggestions
Consolidate the deploy-gating rule and the 'generate-metadata is not a deploy' caveat into a single statement each — they are currently repeated across three to four sections.
Move the non-Go language guidance (SQL placeholder syntax, Bash/PowerShell args, TypeScript/Python preprocessor typing) into a referenced file or trim it, since the skill is declared as Go-only.
Add a short error-recovery step for `wmill script preview` failures (read the error, fix, re-preview) to complete the feedback loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Most content is dense, genuinely non-obvious Windmill platform knowledge (lock hashes, preview vs run semantics, rehash), but there is noticeable repetition — "not a deploy" is stated three times (lines 14, 31, 33) and deploy-gating is repeated four times (lines 15, 19, 26, 45) — plus scope creep: a Go-titled skill carries multi-language guidance (SQL placeholders, Bash/PowerShell args, TS/Python preprocessors). This lands between anchor 2's 'several unnecessary explanations' and anchor 4's 'minor instances of over-explanation'; it is mostly useful material that could be meaningfully tightened. | 3 / 5 |
Actionability | The body gives exact commands with flags (`wmill script preview <script_path>`, `wmill generate-metadata rehash`, `wmill generate-metadata --dry-run`, `wmill resource-type list --schema`) and three complete, executable Go code samples including the required `package inner` declaration, return types, and error handling. Concrete cases are covered (e.g. the `requests 2.31.0 → 2.32.0` version-bump example), matching 'Fully executable; copy-paste ready code or commands'. | 5 / 5 |
Workflow Clarity | The write → preview → generate-metadata → deploy sequence is governed by explicit intent-based decision rules ('If the user says run/try/test with local edits, use script preview') and includes validation checkpoints: --dry-run to inspect stale items before regenerating, and diffing regenerated locks to catch version bumps before deploy. It falls short of anchor 5 only because there is no error-recovery guidance for when a preview fails, and the sequence is spread across thematic sections rather than presented as one ordered flow. | 4 / 5 |
Progressive Disclosure | With no bundle files, the single ~130-line document is well-organized into clearly headed sections (CLI Commands, General Principles, Return Values, Preprocessors, Go) with clearly signaled external pointers (the **Deploying** section in `AGENTS.wmill.md`, the `preview` skill) that are one level deep, not nested. Anchor 4 ('Good structure; most content is appropriately placed; minor organization gaps') fits — the non-Go language guidance inlined in a Go skill is the main content that arguably belongs elsewhere. | 4 / 5 |
Total | 16 / 20 Passed |