Content
65%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 router-style SKILL.md with concrete commands, tool names, and an effective error-code troubleshooting loop. Two material weaknesses hold it back: the bundle is missing the files that its most important routes point to (smartcanvas/, doc/, sheet/ entry files and shell scripts), and destructive operations (recursive node deletion) lack validation checkpoints.
Suggestions
Ship the missing bundle files referenced by the routing table and directory tree — `smartcanvas/entry.md`, `sheet/entry.md`, `doc/entry.md`, `setup.sh`, and `import_file.sh` — or repoint those routes to files that exist, since the top-recommended smartcanvas and sheet paths currently dead-end.
Add a validation checkpoint before destructive operations: e.g., 'Before delete_space_node with remove_type=all, call the node-list tool to show the user which child nodes will be recursively deleted and confirm.'
Trim redundancy to improve token efficiency: drop the full directory tree (the routing table already covers it) and reduce the '常见工作流' section to one-line pointers into references/workflows.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and mostly table/rule driven with no explanations of concepts Claude already knows — the document-type table, scenario routing table, and terse '核心规则' bullets each earn their tokens. It is not a 5 because of redundancy: the full directory tree repeats the routing table's file list, and the '常见工作流' section re-summarizes content already delegated to references/workflows.md; the error table's inline upsell URLs are also padded. | 4 / 5 |
Actionability | Concrete, executable patterns are given throughout: the exact command form `mcporter call "tencent-docs" "<工具名>" --args '<JSON参数>'`, a complete update-check command with JSON response fields, specific tool sequences ('manage.search_file 按关键词搜索 → 获取 file_id → get_content 读取内容'), and named parameters (`parent_id`, `remove_type=current`, `file_id`). It is not a 5 because most invocation examples lack concrete argument payloads — detail is delegated to reference files — and `mcporter list` output inspection is described rather than exemplified. | 4 / 5 |
Workflow Clarity | Sequences exist (search → file_id → get_content; scrape_url → scrape_progress polling; async_import → import_progress) and the 排查步骤 section provides a numbered error-recovery loop keyed to error codes. However the skill includes destructive operations — '删除节点需谨慎:delete_space_node … 使用 `all` 时会递归删除所有子节点' — with only a caution and no validation checkpoint (e.g., list nodes/confirm with user before delete), so per the rubric's cap, a destructive workflow without validation cannot score above 3. | 3 / 5 |
Progressive Disclosure | The structure is well designed — a scenario routing table, a directory tree, and one-level-deep references under references/ — but scored against the actual bundle, several primary navigation targets do not exist: `smartcanvas/entry.md`, `sheet/entry.md`, `doc/entry.md`, `setup.sh`, and `import_file.sh` are all referenced (the smartcanvas and sheet routes are the top-recommended paths) yet missing from the bundle. References that are present are clearly signaled, but broken routing for the highest-priority scenarios means navigation fails exactly where it matters most, fitting score 3 ('could be better organized') rather than 4. | 3 / 5 |
Total | 14 / 20 Passed |