Content
81%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 strong operational skill document: tiered cost-aware retrieval with explicit decision points, guards, fallbacks, and copy-paste-ready commands. Its main weakness is token efficiency — a long single file with some redundant guard restatements and padded explanatory passages that could be trimmed or split into reference material.
Suggestions
Tighten the verbose sections: compress the Visibility Filter's no-op-default explanation, the _raw/ defensive-filter paragraph, and the event-time rationale to their operative rules; state the read-only guard once at the top instead of three times across Steps 5 and 6.
Split detail-heavy reference material (QMD transport/command selection, lifecycle trust-annotation table, structural-intent output fields) into a references/ file with clearly signaled one-level-deep links, keeping SKILL.md as a leaner overview.
Surface one or two of the skill's distinctive answer modes (structural intents like impact/hubs/bridges, temporal --as-of lookups) in the frontmatter description so trigger coverage matches the body's actual capabilities.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is operationally dense and avoids general-knowledge padding, but contains several sections that could be tightened: the Visibility Filter opens by explaining a no-op default at length ('By default, all pages are returned regardless of visibility tags. This preserves existing behavior — nothing changes unless the user asks for it'), the _raw/ defensive-filter paragraph runs long, and the read-only guard is stated three times (top section, Step 5, Step 6). Mostly efficient with some unnecessary explanation — an anchor-3 fit, not 4 (multiple sections need trimming) and not 2 (nothing is generic filler Claude already knows). | 3 / 5 |
Actionability | Fully executable commands throughout: 'obsidian-wiki graph-query "$OBSIDIAN_VAULT_PATH" "<question>" --pretty', 'Grep -A 10 -B 2 "<query-term>" <candidate-file>', '${QMD_CLI:-qmd} query $'lex: ...\nvec: ...' -c "$QMD_WIKI_COLLECTION" -n 8 --files', and the Step 6 log command with exact parseable field names. Concrete grep patterns, frontmatter field names, and output-field semantics cover the common cases copy-paste ready. | 5 / 5 |
Workflow Clarity | The tiered pipeline (Step 0 pre-pass → index pass → QMD → section pass → full read) has an explicit decision tree with stop conditions at every tier ('If graph is non-null → ... stop; no page reads', 'If index_only: true → skip Steps 1-4'), guards (QMD guard, Step 4b cost guard), and error-recovery fallbacks ('if obsidian-wiki is not installed: proceed with Step 1'; unavailable transport → continue with Step 3). Explicit checkpoints plus feedback loops match the top anchor. | 5 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent), and the single SKILL.md is well-sectioned with a clear step-based navigation structure; external references (llm-wiki/SKILL.md for config resolution and retrieval primitives, the $qmd-cli skill, .env.example) are clearly signaled inline and only one level deep. Not 5 because at ~330 lines, detail-heavy sections (QMD transport modes, trust annotation tables, the visibility filter) could live in a separate reference file to keep the main file a leaner overview. | 4 / 5 |
Total | 17 / 20 Passed |