Content
65%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a dense, actionable orientation skill with concrete commands and a clear routing table, but it is somewhat long and its validation feedback loops are implicit. Its referenced detail files are signaled but missing from the bundle, weakening progressive disclosure.
Suggestions
Make the authoring loop's validation explicit with a 'validate, then only proceed when clean' checkpoint (e.g. run vet-placeholders/vet-schemas and stop if they fail) to strengthen workflow clarity.
Either add the referenced files under references/ (cub-cli.md, functions-catalog.md, filters-and-queries.md, triggers-recipes.md, yaml-patterns.md) or remove the dangling References entries so progressive disclosure points to real material.
Tighten repetition of the 'one resource per Unit' and config-as-data tenets into a single authoritative statement to improve conciseness.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~200-line body is mostly necessary product-specific doctrine (no generic-concept padding Claude already knows), but it is long and repeats key tenets ('one resource per Unit', config-as-data) across sections, so it could be tightened. Not level 3 because not every token earns its place; not level 1 because it avoids explaining concepts Claude already knows. | 2 / 3 |
Actionability | It provides concrete, executable commands throughout — e.g. 'cub unit get <example-slug> --space skill-examples -o yaml', 'kubectl create ... --dry-run=client -o yaml | yq \'del(.metadata.creationTimestamp, .status)\'', 'cub space update ... --label', plus gate and promotion commands. Guidance is copy-paste ready rather than pseudocode. | 3 / 3 |
Workflow Clarity | The authoring loop is a clear 1–5 sequence and validation is mentioned (vet-placeholders/vet-schemas/vet-format, 'Confirm before you compose', Verify chain), but the validate→fix→retry feedback loop is implicit rather than the crisp 'only proceed when valid' pattern. It is above level 1 because steps and validation are present, but below level 3 because checkpoints are not explicit. | 2 / 3 |
Progressive Disclosure | Sections are well-organized and a References list signals one-level-deep files (cub-cli.md, functions-catalog.md, yaml-patterns.md, etc.), but those referenced files do not exist in the bundle (references/, scripts/, assets/ are absent), so the disclosure paths are broken and the body remains fairly substantial. Not level 3 because the signaled references are not actually present; not level 1 because structure and signaling are good. | 2 / 3 |
Total | 9 / 12 Passed |