Content
81%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 well-structured, highly actionable skill: all five modes have numbered steps, concrete CLI commands, and explicit validation blocks, and it assumes Claude's Rego competence rather than teaching basics. The main improvements are trimming the wizard prompt duplication and splitting per-mode detail into reference files.
Suggestions
Add one minimal complete example deny rule in the generate mode so the output is copy-paste ready rather than structural guidance.
Compress the wizard's fully quoted prompts into a short table of mode-to-context-question mappings; the prompt text currently duplicates the mode descriptions.
Move the longer per-mode detail (debug checklist, test naming conventions) into references/ files linked from each mode section to keep SKILL.md an overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational, assuming Rego knowledge ('Use `some` for iteration, `in` for membership'), but the wizard's fully quoted prompts duplicate the per-mode descriptions and could be tightened into a compact table — minor instances of over-explanation, matching anchor 4 rather than the lean anchor 5. | 4 / 5 |
Actionability | Concrete commands appear throughout ('conftest verify --policy <dir>', 'regal lint <dir>', "grep -oE 'input\.[a-zA-Z0-9_.[\]]+'"), but the generate mode describes policy structure without a single example rule and the jq filter is a placeholder pattern rather than executable code — mostly executable with minor gaps, short of the copy-paste-ready anchor 5. | 4 / 5 |
Workflow Clarity | Every mode is a numbered sequence with explicit validation checkpoints: generate requires both a passing and a failing resource, test requires 'zero skipped', validate chains fmt --check to auto-format to lint to verify with a per-step report, and debug ends with 'how to verify it fires' — the explicit-validation-and-feedback-loop anchor. The one in-place rewrite (auto-format) is gated behind the fmt --check, so no risky step lacks validation. | 5 / 5 |
Progressive Disclosure | Headers and mode sections are well organized and every command is real, but the ~130-line body keeps per-mode detail (debug checklist, test naming conventions, wizard prompts) inline in a single file with no one-level-deep references, fitting 'good structure; minor organization gaps' rather than the split-content anchor 5. | 4 / 5 |
Total | 17 / 20 Passed |