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-structured instruction-only skill: lean body, concrete tool and command references, explicit backup/validate/verify checkpoints, and clean routing to three real one-level-deep reference files. The main gaps are minor: a few trimmable rationale sentences and the fact that full executable command sequences live entirely in the references rather than the body.
Suggestions
Trim background rationale that restates what Claude already knows (e.g. the OS-differences sentence in Golden rule 1 and the setup-cost justification in rule 2), keeping just the instruction.
Inline one or two complete copy-paste command examples for the most common case (e.g. a full backup → edit → validate JSON → restart Claude Desktop → verify sequence) so the body alone handles the top task without a reference hop.
Move the generic 5-step 'agent isn't responding' diagnostic into a reference file or fold it into the per-tool references, keeping only a one-line pointer in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient and adds skill-specific knowledge (Desktop Commander tool names, failure-mode heuristics) rather than generic concepts, but a few rationale sentences could be trimmed — e.g. "file paths, commands, and config locations differ across macOS, Windows, and Linux" (Claude knows this) and "The most common reason these sessions dead-end is a provider auth or billing error discovered *after* a lot of setup work." This matches anchor 4 ("efficient; minor instances of over-explanation that could be trimmed") rather than anchor 5, where every token earns its place. | 4 / 5 |
Actionability | Concrete, mostly executable guidance throughout: named Desktop Commander tools (`get_config`, `start_process`, `edit_block`, `list_processes`), specific commands (`cp file.json file.json.bak`, `lsof`/`netstat`, `hermes setup --portal`), exact config paths, and docs URLs. However, complete copy-paste command sequences for the common cases are deferred to the reference files rather than present in the body, so anchor 4 ("concrete code or commands with minor gaps") fits better than anchor 5's "copy-paste ready" coverage. | 4 / 5 |
Workflow Clarity | The 6-step general workflow is clearly sequenced and validation is explicit, not implied: back up before editing ("Copy the config to a `.bak` first"), "Validate the JSON after every edit", "use `list_processes` to confirm the process is running and `start_process` (`lsof`/`netstat`) to confirm the port is listening", plus the 5-step ordered diagnostic as an error-recovery checklist and rule 3's "suggest a concrete working alternative instead of retrying". This satisfies anchor 5 (explicit validation steps, feedback loops, checklist); anchor 4's "minor validation gaps" does not describe this body. | 5 / 5 |
Progressive Disclosure | The body is a concise routing overview with clearly signaled, one-level-deep references — each task bullet ends with "→ Read `references/claude-desktop-mcp.md`" (and openclaw.md / hermes.md), and all three files exist in the bundle with no nested `.md` references inside them. Detail (exact paths, commands, failure modes) is appropriately split out of SKILL.md, matching anchor 5 ("clear overview with well-signaled one-level-deep references; easy navigation"). | 5 / 5 |
Total | 18 / 20 Passed |