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 dense, highly actionable CLI skill: copy-paste-ready commands throughout, strong safety validation and feedback loops around destructive and batch operations, and almost no filler. Its one structural weakness is that the whole command reference lives inline in SKILL.md with no bundle files, so the context cost is paid on every invocation.
Suggestions
Split low-frequency command detail (tags, presets, adopt/set-source variants, legacy sync) into references/ files (e.g. references/presets.md, references/sources.md), keeping SKILL.md to the CLI resolution step, mental model, and the common install/deploy/remove workflows.
Trim the safety prose in 'Before doing anything' and the set-source --force section to their operative instructions — the behavior rules are essential but several sentences explain the same refusal twice.
Deduplicate the --agent-required rule and the pitfalls that restate guidance already given in their command sections, keeping the Pitfalls section for cross-cutting failure modes only.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Command blocks dominate and nearly all prose conveys CLI semantics Claude cannot know (BRIDGE_BROKEN resolution, held_back_removals, --force semantics), so it is well past anchor 3. It falls just short of anchor 5 because a few passages are discursive beyond necessity — e.g. "that binary may predate a safety fix, and the machine has a desktop app whose version nothing here can match" — and the --agent-required and pitfalls points are each restated more than once. | 4 / 5 |
Actionability | Every section gives copy-paste-ready commands ("$SM" skills install https://github.com/foo/bar/tree/main/skills/baz, the full CLI-resolution script, exact JSON error shapes), and the $SM placeholder is handled with an executable resolution step plus an explicit warning not to carry it across shells. Concrete examples cover the common cases. | 5 / 5 |
Workflow Clarity | "Typical workflows" sequences steps with explicit validation checkpoints ("Always verify after install" via skills status, --dry-run previews before batch operations, check before update), and destructive operations get real feedback loops (held-back paths shown and escalated to the user, TARGET_CONFLICT paths reported, never auto-confirm removes). The destructive/batch validation cap does not bite — validation is thorough. | 5 / 5 |
Progressive Disclosure | The body is well-structured with clear section headers, but there are no bundle files at all and the entire CLI reference (~350 lines: ref-resolution rules, adopt, set-source, tags, presets, legacy sync, pitfalls) is inlined in SKILL.md. That matches anchor 3 — good structure, but content that clearly belongs in separate reference files is inline — rather than anchor 2 (structure itself is good, not minimal) or anchor 4 (a reference this size should be split out). | 3 / 5 |
Total | 17 / 20 Passed |