Content
93%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 highly actionable and token-efficient with clean structure and well-signaled references. The only gap is that the validation feedback loop is implied rather than an explicit step in the workflow.
Suggestions
Add an explicit feedback step to the Workflow, e.g. after 'cmux-settings validate': 'If unknown keys are flagged, remove or correct them and re-run validate before reporting success.'
Show one validate-failure example inline (unknown key → unset → re-validate) so the recovery path is copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — it states facts ('JSONC', 'file watcher applies changes on save, no restart') without explaining JSON or config concepts, and every section earns its tokens. | 5 / 5 |
Actionability | Provides a full command table plus copy-paste-ready examples like 'cmux-settings set sidebarAppearance.matchTerminalBackground true' and 'cmux-settings set shortcuts.bindings.newTab ["ctrl+b","c"]' covering common cases. | 5 / 5 |
Workflow Clarity | The four-step workflow includes an explicit 'Read back and cmux-settings validate' checkpoint, but the validate→fix→re-validate feedback loop for failed validation is only implied in the Rules section rather than spelled out in the sequence. | 4 / 5 |
Progressive Disclosure | SKILL.md is a concise overview with one-level-deep, clearly linked references (references/all-keys.md, references/shortcut-actions.md) and a real bundled helper script, all verified to exist in the bundle. | 5 / 5 |
Total | 19 / 20 Passed |