Content
85%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-engineered skill body: concrete commands, an audited-then-verified workflow with explicit checkpoints and stop conditions, and disciplined scope and safety rules. The only trims available are repeated guardrail phrasing and a slightly tangential catalog note, and the absence of any sample YAML entry is the main actionable gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense, operational guidance Claude cannot infer (manifest-based auditing, house-style mirroring, scope discipline), with clear sections. It is not a 5 because guardrails repeat across sections — "never commits or pushes" appears in the intro, workflow step 6, and the mistakes table; "no warehouse needed/needed no warehouse connection" appears twice in step 1 — and the catalog.json digression is slightly tangential. It is above 3 because nearly every token is project-specific know-how rather than padding or concepts Claude already knows. | 4 / 5 |
Actionability | Executable commands are given verbatim ("dbt parse", "python3 <SKILL_BASE_DIR>/audit_coverage.py <folder>", "git diff <folder>", "--manifest <path>") and the drafting steps are concrete (identify grain, resolve every ref()/source(), reuse upstream wording, greenfield fallback recipe). It falls short of 5 because the skill's main output artifact — a drafted YAML entry — is never shown by example, even for the greenfield case where no house style exists to mirror; it is above 3 because everything needed to execute is present and specific. | 4 / 5 |
Workflow Clarity | Six clearly sequenced steps with an explicit validation checkpoint and feedback loop: step 5 requires re-running "dbt parse" ("must be clean before handing back") and re-running the audit to confirm the closed gap, and step 1 has an explicit stop condition ("If the audit shows 0 gaps, report full coverage and stop"). The mistakes table adds a red-flag checklist, matching the top anchor for batch operations with validation. | 5 / 5 |
Progressive Disclosure | A single-file skill, cleanly sectioned (conventions, workflow, untrusted input, scope, mistakes), whose only external dependency (audit_coverage.py) is clearly signaled with full invocation instructions, plus one-level-deep pointers to sibling skills (running-dbt-commands, using-dbt-for-analytics-engineering). No inline content clearly belongs in a separate file, and navigation is easy, so it sits at the top anchor rather than 4. | 5 / 5 |
Total | 18 / 20 Passed |