Content
76%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 lean, highly actionable skill body whose documented commands all match the real bundled script. The main weakness is workflow clarity: batch write operations lack validation/verification guidance, and a "Step N" placeholder heading sits with no numbered steps, alongside minor conciseness issues (duplicated description line, pinned model version).
Suggestions
Add validation guidance for batch operations, e.g. run --scan-recent before --scan-all, verify connections.json and the Connections sections after writes, and how to recover from API failures mid-scan.
Fix or remove the "## Step N: Reflection" placeholder heading — there are no Steps 1..N-1 in the body — and consider moving the reflection boilerplate into a short reference or the calling convention.
Trim the duplicated opening line under the title and move the pinned model (grok-4.5) and machine-specific vault path into a clearly-marked configuration note so they do not go stale.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient — most tokens are executable commands and concrete steps — but the opening line "Detects and classifies relationships between Obsidian vault notes using AI-powered analysis" duplicates the frontmatter description with a buzzword, and the pinned model version "grok-4.5" is time-sensitive information outside any deprecated/old-patterns section. Minor trims needed, matching anchor 4 rather than the fully lean anchor 5. | 4 / 5 |
Actionability | The Usage section provides five copy-paste-ready commands ("python scripts/detect.py --note ...", "--scan-recent 5", "--scan-all", "--update-mocs"), all verified against the actual argparse in scripts/detect.py, plus concrete configuration values and file paths. Specific examples cover the common cases, matching anchor 5; the undocumented --no-section flag is a gap only for an uncommon case. | 5 / 5 |
Workflow Clarity | The seven "What it does" steps are clearly sequenced, but the skill performs batch write operations (full corpus scan calling the API for every candidate pair, editing notes and MOCs) with no validation or verification guidance — the rubric's cap for batch operations without validation applies. Additionally, "## Step N: Reflection" is a placeholder heading with no Steps 1 through N-1 preceding it, weakening coherence. Not 2, because a rough-to-clear sequence is present. | 3 / 5 |
Progressive Disclosure | Good structure: the 827-line detection logic is appropriately split into scripts/detect.py, which is referenced with a clear path and exists in the bundle, and sections (What it does, Usage, Configuration, Integration) are well organized. Not 5: the composable reflection boilerplate is oddly placed at the end with a broken heading, and there are no reference docs for detect.py's behavior beyond the code itself. | 4 / 5 |
Total | 16 / 20 Passed |