Content
75%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 SKILL.md body is a well-structured, token-efficient routing index: it gives executable doc-fetch commands, explicit anti-guessing rules, and cleanly organized one-level-deep reference tables. Its main weaknesses are the verbatim duplication of the frontmatter description, generic when-to-use bullets, an implicit rather than explicit workflow sequence, and reference files that could not be verified in the bundle as provided.
Suggestions
Remove the opening paragraph that duplicates the frontmatter description verbatim, and tighten or drop the generic 'When to Use This Skill' bullets — the routing tables already convey this more concretely.
Make the docs workflow explicit, e.g. '1. Pick the resource file from the tables below; 2. Fetch live docs via llms.txt only when the reference file lacks the answer', and include one worked curl example with a real docs path instead of the `<path>` placeholder.
Fix the small copy errors ("additonal" → "additional", "support you find" → "help you find") to match the skill's otherwise polished presentation.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by lean routing tables that earn their tokens, but the opening paragraph repeats the frontmatter description verbatim, the 'When to Use This Skill' bullets are generic (e.g. "Questions about Neon best practices"), and there are minor typos ("additonal", "support you find"). This matches anchor 4 — efficient with minor instances that could be trimmed — rather than 5's every-token-earns-its-place. | 4 / 5 |
Actionability | Concrete executable commands are provided: `curl https://neon.com/llms.txt` and `curl -H "Accept: text/markdown" https://neon.com/docs/<path>`, plus the explicit rule "Don't guess docs pages. Use the llms.txt index". The `<path>` placeholder is unfilled and no worked fetch example is given, so it is 'mostly executable with minor gaps' (anchor 4) rather than fully copy-paste ready for the common case. | 4 / 5 |
Workflow Clarity | The routing workflow is unambiguous: each resource table pairs a file with a 'When to Use' column, and the docs-fetch path is prescribed via llms.txt. No destructive or batch operations exist, so the validation cap does not apply, but the reference-then-fetch-docs sequence is implicit rather than explicitly sequenced with checkpoints — anchor 4, not 5's explicit step/checkpoint structure. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview that routes to 13 well-signaled, one-level-deep reference files organized in labeled tables with per-file 'When to Use' guidance — close to anchor 5's structure. However, none of the referenced `references/*.md` files exist in the provided bundle, so the referenced paths could not be verified as real files, keeping it at anchor 4 rather than 5. | 4 / 5 |
Total | 16 / 20 Passed |