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 an exceptionally lean, well-organized command doc: a clear six-step workflow, concrete paths and citation formats, and disciplined delegation to a sub-agent and reference files. The gaps are modest — no executable invocation examples for the two scripts, no handling for the no-results case, and references that point outside the skill's own bundle.
Suggestions
Show a one-line invocation example for each script (e.g. how to pass a query to wiki_search.py and what it returns) so the fallback and logging steps are copy-paste ready.
Add a brief branch for the zero-relevant-pages case (e.g. "if the index and BM25 both come up empty, say so and suggest filing the question as a stub page") to close the workflow's validation gap.
Since the referenced files (query-workflow.md, agents/wiki-librarian.md) are not part of this skill's bundle, consider either shipping them as local bundle files or stating in the body that this is a thin root-level copy that depends on the llm-wiki skill's files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body never explains concepts Claude already knows (BM25 is named, not taught; sub-agents and wikilinks are used without preamble) and every section is functional: usage examples, a 6-step process, an output-format table, script paths, and three rules. Not 4: there is no over-explanation or padding to trim — the mild overlap between the usage examples and the output-format table serves different purposes (invocation vs. output shape). | 5 / 5 |
Actionability | Concrete guidance throughout: "reads 3-10 pages in full", "runs engineering/llm-wiki/skills/llm-wiki/scripts/wiki_search.py (BM25)", "inline [[sources/xxx]] citations", "usually in comparisons/ or synthesis/". Not 5: the scripts are named by path but no invocation example is given (e.g. the argument shape for wiki_search.py), so the fallback and logging steps are not copy-paste ready; not 3: the guidance is specific and executable in structure, not pseudocode. | 4 / 5 |
Workflow Clarity | The "What happens" section gives a clear 6-step sequence (index read → drill-in → follow links → fallback search if the index isn't enough → synthesize → offer file-back), with the fallback acting as a conditional checkpoint and the file-back gated on user confirmation. Not 5: there is no explicit validation of the synthesized answer and no guidance for the zero-relevant-pages case; not 3: the sequence is well defined with a conditional recovery path and the Rules section enforces per-claim citations as a quality check. | 4 / 5 |
Progressive Disclosure | The body is a lean overview that delegates detail via clearly signaled one-level-deep references ("agents/wiki-librarian.md", "→ engineering/llm-wiki/skills/llm-wiki/SKILL.md", "→ engineering/llm-wiki/skills/llm-wiki/references/query-workflow.md") with a dedicated Skill Reference section. Not 5: the referenced files live outside this skill's directory (no local references/, scripts/, or assets/ bundle ships alongside this SKILL.md), so navigation depends on the wider repo layout and the repo-root-relative paths noted in the sync comment; not 3: structure is good and references are explicit, not buried. | 4 / 5 |
Total | 17 / 20 Passed |