Content
57%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.
A dense, accurate, and highly actionable CLI reference that assumes Claude's competence and documents product-specific flags Claude cannot know. Its weaknesses are structural: a monolithic ~900-line single file with no reference files, duplicated option blocks, and no explicit validation guidance for the many destructive push/merge/delete operations it documents.
Suggestions
Split the per-command reference into per-domain reference files (e.g. references/jobs.md, references/sync.md, references/apps.md) and keep SKILL.md as a concise overview with one-level-deep pointers, per the progressive-disclosure model.
Deduplicate the `sync pull`/`sync push` and `instance pull`/`instance push` option blocks by stating shared options once, and factor the ~40 repeated "--json - Output as JSON (for piping to jq)" lines into a single global note.
Add validation guidance for destructive/batch operations — e.g. "run `wmill sync push --dry-run` (or `--lint`) and review the changes before pushing without `--yes`" — so workflow clarity isn't capped by missing checkpoints.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Entries are terse one-line flag descriptions with no conceptual padding (e.g. "--show-diffs - Show diff informations when syncing"), and the "Key concepts (not obvious from per-command --help)" section adds genuinely non-obvious guidance. However, `sync pull`/`sync push` repeat ~28 identical options, `instance pull`/`instance push` repeat ~10, and "--json - Output as JSON (for piping to jq)" is repeated ~40 times — more than minor over-explanation, matching anchor 3 ("could be tightened") rather than 4. | 3 / 5 |
Actionability | Exact command signatures with argument types, flags, and defaults throughout (e.g. "object-storage upload <local_path:string> <file_key:string>", "-d --data <data:string>" with explicit $res/$var semantics) let Claude construct commands directly. It falls short of anchor 5 only because there are few complete worked examples with real values or sample outputs. | 4 / 5 |
Workflow Clarity | This is a lookup reference rather than a sequenced workflow; implicit sequencing exists (the object-storage "Choosing a subcommand" decision guide, digest's "Run from the root of a sync checkout" timing note). But destructive/batch operations — "sync push" ("This overrides any remote versions"), "workspace merge", "object-storage delete", "datatable migrate down" — lack explicit validation checkpoints even though the tools exist (--dry-run, --lint, --locks-required), so the rubric cap of 3 applies. | 3 / 5 |
Progressive Disclosure | No bundle files exist; the entire ~840-line command-by-command API reference is inlined in SKILL.md. Internal structure is good (## Global Options, ### per command group, a separate # Object Storage CLI section), making it navigable — matching anchor 3 ("some structure... content that should be separate is inline") rather than 2, which requires minimal structure or no headers. | 3 / 5 |
Total | 13 / 20 Passed |