Conciseness

The body is dense and almost entirely tool-specific knowledge Claude cannot know (the ':from:count' suffix, --full-path semantics, the filter/match predicate AST, line-numbered output headers), with no padding about general concepts. It matches anchor 4 rather than 5 because the 'write intent:/lex:/vec: yourself — do not lean on query expansion' exhortation is repeated in three or more places and could be consolidated into one.

Actionability

Fully executable, copy-paste-ready commands throughout: search/query/multi-get invocations with flags, a wrong/right pair for line slicing ("Wrong: qmd get '#abc123' | sed -n '120,160p'" / "Right: qmd get '#abc123:120:40'"), complete filter-AST JSON examples, the MCP query payload, and troubleshooting commands — specific examples cover the common cases, matching the top anchor.

Workflow Clarity

An explicit numbered core workflow ("The workflow is always: 1. Search for candidate documents. 2. Retrieve the full source... 3. Answer from retrieved text, citing paths or docids") reinforced by a typical loop, with explicit error-recovery feedback: fall back to 'qmd search' when model/GPU setup fails, run 'qmd doctor' before changing configuration, use '--explain' to inspect ranking, and check metadata coverage before filtering. The Pitfalls section acts as a checklist; core operations are read-only so the destructive-operation cap does not apply.

Progressive Disclosure

Well-sectioned body with a clearly signaled, verified, one-level-deep reference ("See references/mcp-setup.md") that exists in the bundle and contains no nested references. It matches anchor 4 rather than 5 because the 355-line SKILL.md is a full usage manual — roughly 90 lines of metadata filter/discovery detail and the MCP 'query' tool section (which overlaps references/mcp-setup.md) could be split into reference files.

Specificity

The description explicitly names several concrete actions — "Search local markdown knowledge bases, notes, docs, and wikis", "find notes, retrieve documents, inspect a wiki, answer from indexed markdown, or set up QMD access" — which matches the anchor for several specific actions with minor gaps. It is not a 5 because the 'what' compresses capability under the single verb 'Search' (retrieval, citation, and query-structuring detail live in the body), and not a 3 because more than 1-2 concrete actions are explicitly stated.

Completeness

It clearly answers both questions: what ("Search local markdown knowledge bases, notes, docs, and wikis with QMD") and when ("Use when users ask to find notes, retrieve documents, inspect a wiki, answer from indexed markdown, or set up QMD access") with concrete trigger phrases, structurally matching the anchor-5 example. The 'when' clause is explicit, not merely implied.

Trigger Term Quality

Natural phrases users would say are present: "find notes", "retrieve documents", "inspect a wiki", "wikis", "docs", "knowledge bases", "indexed markdown". Good keyword coverage with a few natural terms missing — no file extensions (.md) and no common synonyms like 'second brain' or note-app names — so it matches anchor 4 rather than the comprehensive synonym-plus-extension coverage of anchor 5.

Distinctiveness Conflict Risk

A clear niche — searching indexed local markdown collections via the named QMD tool — with distinct triggers, but the generic nouns "notes", "docs", and "documents" leave minor overlap risk with generic documentation-search or note-management skills (anchor 4). It lacks the fully minimal conflict risk of anchor 5, while being far more distinct than anchor 3's 'could still overlap with similar skills'.