Content
36%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 a catalogue advertisement rather than a usable skill: it duplicates the frontmatter and defers all real guidance to an unbundled external repository, with the agent told to open a browser to find the actual instructions. Nothing in the file is executable toward the skill's stated purpose. Either bundle the upstream material with concrete entry points, or replace the deferral with a specific, copy-paste-ready install and invocation sequence.
Suggestions
Replace `open https://github.com/openai/skills` with the concrete install steps — e.g. the exact clone/copy commands and target skills-directory path — so the agent can act without browsing.
Drop the "What it does" and "Source" sections that restate the frontmatter description and upstream URL verbatim, and use the space for a minimal quick-start example of the workflow.
If the upstream bundle is the real content, vendor it (or its key references) into references/ with explicit paths, so the one-level-deep pointer resolves to actual files rather than "inspect the README".
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is short, but the "What it does" and "Source" sections restate the frontmatter description and upstream URL verbatim, so tokens are spent on duplication rather than new information. Not 4: the repetition is unnecessary padding; not 2: there is no extensive over-explanation of concepts Claude already knows. | 3 / 5 |
Actionability | The only concrete command is `open https://github.com/openai/skills`, which tells the agent to manually browse a repository rather than execute a step; there is no install command, no paths, and no example of invoking the skill. Not 3: even pseudocode-level guidance toward the actual task is absent; not 1: it does point to a specific URL and names the skill to invoke. | 2 / 5 |
Workflow Clarity | A rough sequence exists (inspect upstream README → install the bundle → invoke by name or trigger phrase), but every step is underspecified: how to install, into which directory, and what to verify afterward are all missing. Not 3: the gaps go beyond missing validation checkpoints to undefined steps themselves; not 1: the three-step order is present and coherent. | 2 / 5 |
Progressive Disclosure | Sections are clearly labeled and the deferred content is one level deep (the upstream repo), but no references/, scripts/, or assets/ bundle files exist, and navigation is delegated to "Inspect the upstream README for exact paths" — a buried reference the agent must discover itself. Not 4: the pointer to the real material is vague rather than clearly signaled; not 2: the file itself is well-structured and short rather than an inlined wall of text. | 3 / 5 |
Total | 10 / 20 Passed |