Content
77%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 highly actionable with executable commands for every subcommand, clear sequenced workflows, explicit validation for batch enrichment, and an auth-error recovery loop. Its weaknesses are redundancy (the prefer-native-tools guidance repeated several times and usage patterns that restate section content) and a monolithic single-file structure where per-subcommand reference material could be split into bundled files.
Suggestions
Consolidate the "prefer Hermes native" guidance into one place (e.g. "When to use it") and remove the repetitions in the intro, core rule set, and Pitfalls.
Move per-subcommand detail (FindAll, Monitor, Enrichment flag reference) into references/ files (e.g. references/findall.md) and keep SKILL.md as an overview with one-level-deep pointers.
Trim "Recommended Hermes usage patterns" to only workflows not already covered step-by-step in their own sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient command examples with no explanations of known concepts, but the "prefer Hermes native tools" guidance is repeated roughly four times (intro expectations, "When to use it", core rule 7, and Pitfalls) and "Recommended Hermes usage patterns" restates guidance already given in each section. It is not 4 because this redundancy is more than minor trimming; not 2 because there is no padded conceptual explanation. | 3 / 5 |
Actionability | Every subcommand (search, extract, research, enrich, findall, monitor) has copy-paste-ready executable commands with concrete flags and realistic examples, plus stdin patterns and a documented exit-code table. This matches the fully executable, common-cases-covered anchor. | 5 / 5 |
Workflow Clarity | Multi-step workflows are explicitly sequenced (launch with --no-wait, capture the returned ID, continue other work, poll, summarize with citations), the batch enrichment workflow includes an explicit validation checkpoint ("Validate the output file before reporting success"), and the error-handling section provides a feedback loop for auth errors (check auth, confirm key/login, verify PATH). | 5 / 5 |
Progressive Disclosure | The skill is a single ~380-line file with no bundle files at all; all subcommand API detail (quick reference tree, per-command flags, FindAll/Monitor/Enrichment reference material) is inlined in SKILL.md under good section headers. This matches the anchor of well-structured content where a sizable API reference that could live in separate files is inline. Not 4 because no references exist and the file well exceeds the under-50-lines simple-skill exception. | 3 / 5 |
Total | 16 / 20 Passed |