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.
An exceptionally actionable, well-sequenced CLI skill: exact commands, flags, exit codes, error-recovery branches, and a clean overview-plus-guides structure with all 11 referenced files real and one level deep. The one drag is conciseness — Critical Rule 1's per-command envelope enumeration and the duplicated routing tables inline material that the per-family guides already own.
Suggestions
Split Critical Rule 1: keep the universal envelope facts (`Result`/`Code`/`Data`, PascalCase keys, failure-envelope shape) inline and move the per-command `Instructions`/`Pagination` enumeration into the guides that own those commands, which already list their own output shapes.
Trim the Scope Boundaries table of rows that duplicate Task Navigation and the description's routing (e.g. dashboard and Looker rows that just re-link the same guides), keeping only the true out-of-skill redirects.
Consider condensing the generic 'Reasoning budget' and 'Working style' sections to the few rules that are specific to this skill's commands, since the rest restate general agent behavior.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body never explains concepts Claude already knows and is information-dense, but Critical Rule 1 is a single ~250-word sentence exhaustively enumerating which commands add `Instructions` and `Pagination` — detail that belongs in the per-family guides — and the Scope Boundaries table duplicates routing already stated in Task Navigation and the description. This matches anchor 3 ("mostly efficient but includes some... content that could be tightened") better than anchor 4's "minor instances", because the Rule 1 enumeration is a substantial inline block, not a trimmable sentence or two. | 3 / 5 |
Actionability | Guidance is fully executable: exact commands with literal flags (`uip login status --output json`, `--time-range <minutes>` with 60/1440/10080/43200 conversions, `--started-after`/`--started-before` in epoch milliseconds), a table mapping each question to a named playbook with its flags and defaults, and exit codes plus envelope keys for every failure mode. This matches anchor 5 (copy-paste ready commands covering the common cases) with no gaps — flag details beyond the overview are deliberately and correctly delegated to the guides. | 5 / 5 |
Workflow Clarity | The Shared Workflow is a numbered sequence (login check → read the named guide → run the playbook → parse `Data`/`Pagination`) with explicit validation and feedback loops: local ValidationError refusals with exit codes, branching on `Retry` (`RetryWillNotFix` vs `RetryLater`), and Rule 18's read-before-write, `--expected-version`, read-back-after-unknown_error, and mandatory `--backup-file` before any delete. This is anchor 5 — clear sequence, explicit validation, and error-recovery loops — including for the destructive dashboard writes. | 5 / 5 |
Progressive Disclosure | The body is an overview that delegates all command detail to 11 one-level-deep reference guides, each verified to exist in references/ with substantive content, all clearly signaled by name in the Task Navigation table with a stated purpose per guide and explicit read-ordering guidance ("Read only the guides the task needs"), plus rules for when no guide is needed. This matches anchor 5 exactly: clear overview, well-signaled flat references, easy navigation. | 5 / 5 |
Total | 18 / 20 Passed |