Content
86%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, mostly lean workflow skill with highly executable guidance and excellent structure. The main weakness is minor repetition of the same constraints across sections rather than a single canonical statement.
Suggestions
De-duplicate the 'do NOT determinize scoring' rule and the local-first/scope-tightly/name-only constraints — state each once and reference it.
Add an explicit validate→fix→retry loop for the machine-checked output blocks (e.g., re-emit if the block fails format checking).
Consider consolidating the 'Important' recap into cross-references to the relevant steps to trim the recap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely lean with concrete bash and minimal concept explanation, but the 'do NOT determinize scoring' point and the local-first/scope-tightly rules are repeated across the Determinism section, Steps, and the 'Important' recap. | 4 / 5 |
Actionability | Fully executable: copy-paste bash for repo scanning, lib function calls, glab/gh queries, clone plus setup hints, and pinned machine-checked output blocks covering the common GitLab/GitHub cases. | 5 / 5 |
Workflow Clarity | A clear 6-step sequence with decision branches and confirmation gates for the destructive clone, but no explicit validate→fix→retry feedback loop on the machine-checked output blocks. | 4 / 5 |
Progressive Disclosure | No bundle files exist; the only external references (the shared determinism lib and CONTRACT.md) are one-level-deep and clearly signaled, and the body is cleanly sectioned for easy navigation. | 5 / 5 |
Total | 18 / 20 Passed |