Content
82%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 strong, executable reference: concrete commands throughout, a clear sequenced auth workflow, honest scope-setting (what is live vs. what needs current docs), and bundle scripts that match the referenced paths. Its weaknesses are repetition of the docs URL and the duplicated API-key storage command, plus inline detail (option table, drive-share protocol) that would be better split into a reference file.
Suggestions
Deduplicate the docs URL (stated three times) and the API-key storage command (stated twice) to reclaim tokens.
Move the publish.sh option table and the drive-share token protocol into a bundled references file, keeping SKILL.md as the overview.
Add one inline error-recovery note for the publish flow (e.g. what to do when finalize fails) instead of deferring all error handling to the live docs.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is imperative and code-forward with almost no explanation of concepts Claude already knows, matching 'efficient; minor instances of over-explanation that could be trimmed'. The docs URL appears three times (lines 30, 215, 217) and the API-key storage command is repeated verbatim (lines 123 and 155), which is redundant padding. | 4 / 5 |
Actionability | Every workflow ships copy-paste-ready bash with concrete placeholders (e.g. 'bash "$PUBLISH" {file-or-dir} --client hermes'), complete curl calls with real JSON bodies for the auth flow, a real state.json example, and a full flag table for publish.sh. This matches 'fully executable; copy-paste ready code or commands; specific examples cover the common cases'. | 5 / 5 |
Workflow Clarity | The API key flow is a clear numbered 5-step sequence, and the publish flow states its sub-steps ('create/update -> upload files -> finalize. A site is not live until finalize succeeds') plus a defined feedback channel ('Read and follow publish_result.* lines from script stderr'). Not 5 because explicit error-recovery steps for a failed publish are deferred to the live docs rather than given inline, so a checkpoint is implied rather than fully specified. | 4 / 5 |
Progressive Disclosure | Both bundle scripts (scripts/publish.sh, scripts/drive.sh) exist, are referenced correctly via ${HERMES_SKILL_DIR} paths one level deep, and advanced topics (domains, payments, forking, proxy routes) are clearly delegated to a single live docs URL. This is 'good structure; most content is appropriately placed' with a minor gap: the full publish.sh option table and Drive-token handoff detail could live in a bundled reference file instead of the main body. | 4 / 5 |
Total | 17 / 20 Passed |