Content
88%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 high-quality, highly executable skill body: every step is a concrete command, the pipeline is clearly sequenced with an explicit repair/validation pass and a thorough troubleshooting table, and heavy syntax references are properly pushed out to on-demand fetches. The only weaknesses are minor: a few trimmable explanatory clauses and some platform-specific discovery chains that would sit better as bundled scripts.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational — pipeline, flag tables, and platform-specific fallback chains with almost no concept explanation Claude already knows. A few tokens could be trimmed (e.g., explaining that backticks in bash are command substitution, and the 'Inline viewer first' section restating the description's MCP exception), placing it just below 'every token earns its place'. | 4 / 5 |
Actionability | Fully executable throughout: exact CLI invocations with all flags (`drawio -x -f xml -o diagram.drawio diagram.mmd`), complete copy-paste Node/PowerShell/bash one-liners for URL generation and CLI discovery, an XML skeleton, and a working repair-pass script. Concrete examples cover the common cases (Mermaid vs XML, each output format, each platform). | 5 / 5 |
Workflow Clarity | A numbered 3-step pipeline (Author → Deliver → Open) with an explicit repair pass ordered 'right after writing XML... before any layout, export or URL step', a failure fallback ('If the open command fails, print the absolute path'), and a troubleshooting table giving cause→solution feedback loops (e.g., retry with --disable-gpu, two-step Mermaid conversion). Validation and error recovery are explicit, matching the top anchor. | 5 / 5 |
Progressive Disclosure | Good structure: SKILL.md stays an operational overview, delegating Mermaid syntax and XML detail to clearly signaled, conditionally fetched references ('Only when a diagram needs layers, tags, metadata...'), and shipping a real one-level-deep script (scripts/find-shapes.js, verified to exist and match its description). Minor gaps keep it below 5: the inline find-drawio.ps1 PowerShell script and the WSL2/Windows registry-scan chains are bundle-file material kept in the main body, and the syntax references are remote URLs rather than bundled files. | 4 / 5 |
Total | 18 / 20 Passed |